The Treaty on the Limitation of Anti-Ballistic Missile Systems, signed by the United States and the Soviet Union in May 1972, is one of the most studied and least intuitive agreements in the history of #arms_control. It did not limit weapons that kill. It limited weapons that defend. By capping the number of interceptors each side could deploy against incoming warheads, the treaty deliberately preserved each country's vulnerability to the other's nuclear forces. This article explains why two hostile superpowers chose mutual defencelessness, how the treaty was designed and policed for thirty years, why the United States withdrew from it in 2002, and what the world has learned since. Using qualitative document analysis of treaty texts, government reviews, and recent peer-reviewed scholarship, the study argues three points. First, the treaty was not a moral statement but an engineering solution to a specific problem: the interaction between offensive and defensive forces produces an unstable feedback loop unless one side of the equation is fixed. Second, the treaty's institutional design, especially its verification arrangements and its permanent consultative body, was as important as its numerical limits. Third, the arguments that ended the treaty have not been settled by events; they have been reopened. The paper concludes that the central lesson of the #ABM_Treaty is not that #missile_defense is always wrong, but that defence and offence cannot be governed separately, and that any future agreement between nuclear-armed states will have to face this interrelationship directly rather than assume it away. Keywords: Anti-Ballistic Missile Treaty; strategic stability; nuclear deterrence; missile defence; arms control; security dilemma; SALT I; great power competition 1. Introduction 1.1 The problem in plain terms Most people assume that defending your own cities is a peaceful act. Building a shield seems less threatening than building a sword. During the 1960s, a group of American and Soviet officials, engineers, and academics reached the opposite conclusion, and the conclusion was strange enough that it still surprises students today. The reasoning went like this. Two countries each hold enough nuclear weapons to destroy the other. Neither attacks, because attacking means suicide. This grim balance is what the literature calls #mutual_assured_destruction. Now imagine one country builds a #missile_defense system good enough to stop most incoming warheads. That country is no longer suicidal if it strikes first. It could launch, absorb the weakened reply, and survive. The other country, watching this shield rise, has two choices: build more warheads so that enough of them get through anyway, or strike before the shield is finished. Both choices make war more likely. The shield, meant to protect, ends up threatening. This is the logic that produced the 1972 Treaty on the Limitation of Anti-Ballistic Missile Systems, usually called the #ABM_Treaty. It restricted anti-ballistic missile systems on both sides to a token level, and in doing so it locked in the vulnerability of both populations. It was, in the words of a generation of diplomats, the cornerstone of #strategic_stability. 1.2 Why the topic still matters It would be easy to treat the treaty as a museum piece. The United States withdrew from it in June 2002. The Soviet Union that signed it no longer exists. Yet almost every live argument in nuclear policy today is an argument that the #ABM_Treaty tried to settle. The United States is now pursuing a large homeland defence architecture. Russia has fielded weapons whose stated purpose is to fly under, over, or around interceptors. China has expanded its arsenal and is testing its own strategic interception capability. The last bilateral treaty limiting deployed strategic warheads has lapsed. Each of these developments is, in one way or another, an echo of the same #offense_defense_interrelationship that negotiators confronted in Helsinki and Vienna more than fifty years ago. Students of international relations therefore encounter the treaty twice: once as history, and once as a live template for problems that have not gone away. 1.3 Research questions This article addresses five questions. What strategic reasoning led two adversaries to accept limits on their own defences? What did the treaty actually say, and how were its limits made verifiable? How was it implemented, stressed, and reinterpreted between 1972 and 2001? Why did the United States withdraw, and what legal and political arguments were made? What has happened since, and what does the record tell us about the relationship between #ballistic_missile_defense and #strategic_stability? 1.4 Contribution The article makes three contributions. It offers a synthesis suitable for advanced undergraduate and postgraduate students who need the technical, legal, and political dimensions in one place. It connects the 1972 bargain to recent scholarship on the security dilemma, technological drift, and tripolar nuclear competition. And it evaluates, rather than merely restates, the competing claims about what the treaty accomplished and what its removal cost. 2. Conceptual Framework 2.1 Strategic stability and its two parts #strategic_stability is a compound idea. It is usually broken into two components. Crisis stability. In a confrontation, neither side feels pressure to shoot first. If both sides know their nuclear forces will survive an attack, nobody needs to rush. #crisis_stability is high when waiting is safe. Arms race stability. Neither side feels pressure to build more weapons in order to stay safe. #arms_race stability is high when a rival's new programme does not force an expensive answer. The treaty was aimed at both. By capping defences, it protected each side's ability to retaliate, which supported #crisis_stability. By removing the reason to build extra warheads to overwhelm a shield, it supported arms race stability. Recent conceptual work continues to use exactly this two-part structure, and continues to identify homeland defence as one of the strongest single pressures on it (Bolt, 2025). 2.2 The security dilemma The classical #security_dilemma describes a situation where measures one state takes to feel safer make another state feel less safe, even when nobody intends aggression. Defensive weapons are supposed to be the escape from this trap: if a weapon can only defend, it should not frighten anyone. Nuclear defences break that rule. An interceptor cannot destroy a city. But an interceptor placed on top of a large offensive arsenal changes what that arsenal can do. It converts a retaliatory force into a potential #first_strike force, because it promises to mop up whatever the victim manages to fire back. Defence, in the nuclear context, is offence-enabling. This is why the label "defensive" is analytically useless here, and why sober analysts on both sides in the 1960s treated ABM systems as destabilising rather than reassuring. Contemporary scholarship shows the same mechanism at work in Asia. Chinese debate about nuclear expansion is driven substantially by the fear that American defences, precision conventional strike, and cyber capabilities together could erode China's ability to retaliate, producing what has been described as an entangled #security_dilemma in which conventional and nuclear worries feed each other (Hiim, Fravel and Troan, 2023). 2.3 The offence-defence interrelationship The single most important concept in this article is the #offense_defense_interrelationship. Offensive and defensive strategic arms are not two separate policy areas. They are two halves of one equation. If side A doubles its interceptors, side B can restore its confidence by increasing warheads, adding decoys, or shortening flight times. If side B doubles its warheads, side A can respond with more interceptors. The equation has no natural stopping point. It can only be stopped by agreement, and an agreement that limits one half while ignoring the other will not hold. This is not an abstract claim. It is written into the treaty record. It reappeared in the 2010 New Strategic Arms Reduction Treaty, whose preamble acknowledges the link between strategic offensive and strategic defensive arms, and it is the reason that Russian negotiators have repeatedly tied further offensive cuts to limits on American defences. 2.4 Technology creep A fourth concept helps explain what happened after 2002. #technology_creep refers to the steady accumulation of small technical improvements that, taken one at a time, look harmless, but that together produce a capability nobody explicitly decided to build. Sankaran (2025) applies this framework to American #missile_defense. The argument is that radars became far more sensitive than their own requirements demanded, that ship-based interceptors originally built for regional defence were tested successfully against intercontinental targets, and that sensors in space were pushed toward continuous tracking. Each step had a local justification. The cumulative result, in the eyes of Moscow and Beijing, is a system that could be reconfigured against them, whatever Washington says its intentions are. The concept matters because it shows that hostile capability can emerge without hostile intent, and that #arms_control is partly a way of disciplining engineers as well as generals. 2.5 Damage limitation and the theory of the nuclear revolution A fifth concept ties the others together. Damage limitation is the attempt to reduce how much harm an adversary can inflict if deterrence fails. It has two instruments. The first is counterforce: destroying the enemy's launchers, submarines, and command centres before they can be used. The second is active defence: intercepting whatever survives and is fired. The two instruments only work together. Counterforce alone leaves a surviving force that can still devastate you. Defence alone is overwhelmed by a full arsenal. But counterforce followed by defence is a coherent, if terrifying, strategy. It is the reason that even a modest interception capability worries the other side. The question a Russian or Chinese planner asks is not whether an American system could stop a full retaliatory strike, because it obviously could not. The question is whether it could stop the ragged remnant of a force that has already been hit. The theory of the nuclear revolution holds that damage limitation between large nuclear powers is impossible, that everyone knows it, and that this shared knowledge produces caution, restraint, and a certain indifference to the details of the military balance. If the theory is correct, then arms racing is irrational, and the 1972 treaty simply wrote a physical fact into law. The theory has always had critics, and the critics have grown louder. States do not behave as if the balance is irrelevant. They compete over accuracy, over survivability, over warning time, and over defences, which suggests that policymakers do not fully believe that damage limitation is hopeless, or at least are unwilling to bet on it. The persistence of missile defence programmes, in the face of decades of disappointing test results, is itself evidence that the pull of damage limitation is strong, and perhaps stronger than the pull of theory. For the purposes of this article, one does not have to settle the dispute. It is enough to observe that both the believers and the doubters end up in the same place on the practical question. If damage limitation is impossible, defences are a waste of money that frightens others for nothing. If it is possible, defences are the most destabilising thing a state can build. Either way, unconstrained defences are hard to justify. 3. Methodology This study uses qualitative document analysis within a historical-institutional frame. It is not an experimental or statistical design, and it makes no causal claims of the kind that would require such a design. Sources. The evidence base consists of three layers. The first is the primary legal layer: the treaty text and its protocol, agreed statements, and the formal instruments of withdrawal. The second is the official policy layer: posture reviews, defence budget documents, congressional testimony, and government studies, drawn on mainly as reported and analysed in the peer-reviewed literature. The third is the scholarly layer: recent journal articles and monographs, weighted toward work published within the last five years so that the analysis reflects the current state of debate. Selection. Literature was selected on three criteria: peer review or equivalent editorial rigour; direct engagement with the interaction between defences and deterrence; and recency, with priority given to material published from 2021 onward. Older classics are discussed conceptually where they remain foundational, but the citation base is deliberately contemporary. Analysis. The material was coded into five themes that structure Sections 4 to 8: strategic reasoning, legal design, implementation and compliance, withdrawal, and consequences. Competing interpretations were retained rather than resolved where the scholarly record is genuinely divided, in particular on whether the treaty caused the later collapse of arms control or merely preceded it. Limitations. Three limitations should be stated at the outset rather than hidden at the end. Russian-language and Chinese-language primary sources are used only through secondary reporting, which introduces a filter. Much technical information about interceptor performance remains classified, so judgements about effectiveness depend on published test records and on independent scientific assessments. And counterfactual claims, such as what the nuclear balance would look like had the treaty survived, cannot be tested. 4. Historical Background: How the Superpowers Talked Themselves Into Restraint 4.1 The dream of the shield Defence against ballistic missiles was pursued from the moment ballistic missiles existed. The problem was always the same and remains the same: an #ICBM warhead re-enters the atmosphere at several kilometres per second, is small, and can be accompanied by decoys, chaff, and other #countermeasures. Hitting it is possible. Hitting all of them, reliably, in the middle of a nuclear attack, with no rehearsal, is a different matter. Both superpowers tried anyway. The Soviet Union began deploying a system around Moscow, known in Western reporting as #Galosh, in the 1960s. The United States moved through a series of programmes, from Nike-Zeus to Sentinel and then #Safeguard, the last of which was reconfigured to protect missile fields rather than cities. 4.2 The intellectual turn The turning point was conceptual rather than technical. American defence officials, most prominently Secretary of Defense Robert McNamara, began to argue in the mid-1960s that an ABM race would be expensive, technically dubious, and dangerous. Expensive because interceptors cost more than the extra warheads needed to swamp them. Dubious because #countermeasures are cheap. Dangerous because a shield invites a pre-emptive strike against the side that has not finished building one. At the 1967 Glassboro meeting, the Americans put a version of this argument to Soviet Premier Alexei Kosygin, who initially rejected it. His instinctive position was the common-sense one: defending people is moral, threatening them is not. It took several years, and considerable internal Soviet analysis, for Moscow to accept that in the nuclear age this common-sense position produces an uncommon amount of danger. 4.3 Domestic politics on both sides It is a mistake to present the treaty as a triumph of pure reason. Domestic politics did much of the work. In the United States, the Sentinel and #Safeguard programmes ran into fierce local opposition, congressional scepticism, and severe doubts from the scientific community about whether the system would function. Funding was precarious. A treaty that limited Soviet defences while relieving Washington of an unpopular and doubtful programme was politically attractive. In the Soviet Union, the calculation ran the other way. Soviet planners understood that the United States had greater technological depth and could win an unrestrained defensive race, and that American #MIRV technology could saturate Soviet defences faster than Soviet industry could expand them. Both sides, in other words, arrived at restraint partly through fear of losing a race they did not want to run. This is a general feature of successful #arms_control and one of the most useful lessons for students: agreements are usually signed not when states become idealistic, but when both of them expect to lose. 4.4 SALT and the package deal The negotiations that produced the treaty were part of the Strategic Arms Limitation Talks. The outcome, signed in Moscow in May 1972, was a pair of instruments: an Interim Agreement freezing certain offensive launchers, and the #ABM_Treaty limiting defences. The pairing was not accidental. It was the practical embodiment of the #offense_defense_interrelationship. SALT I did not reduce offensive arsenals, and it was widely criticised for that. Its defenders argued, then and now, that capping defences was the precondition for reductions later, because states will not cut their retaliatory forces while an opponent is building a shield. 5. What the Treaty Actually Said 5.1 The core prohibition The heart of the agreement is the undertaking not to build a nationwide defence of national territory against strategic ballistic missiles, and not to provide a base for such a defence. Everything else in the text exists to make that undertaking real. 5.2 Numerical limits Each party was allowed two ABM deployment areas: one around the national capital, and one around an #ICBM launcher field. The two sites had to be separated by a substantial distance, roughly 1,300 kilometres, so that they could not be merged into a regional or nationwide screen. Each site was capped at 100 launchers and 100 #interceptors, along with strictly limited radar deployments. A 1974 Protocol cut the allowance from two sites to one. Each side chose. The Soviet Union kept its capital defence around Moscow. The United States kept the #Safeguard site at Grand Forks, North Dakota, protecting a missile field, and then, in a decision that says a great deal about the perceived value of the technology, shut it down after only a few months of operation. The numbers deserve a moment of reflection. One hundred interceptors, against arsenals numbering in the thousands of warheads, is not a defence. It is a demonstration. The limit was set at a level where the system was strategically meaningless, which was precisely the point. 5.3 The technical fences A treaty that only counted launchers would be trivially easy to evade. The drafters therefore built a set of fences around the edges of the technology. No mobile or exotic basing. The parties agreed not to develop, test, or deploy ABM systems or components that are sea-based, air-based, #space_based_interceptors, or mobile land-based. This provision is the most consequential single sentence in the treaty for present purposes, because almost every contemporary defensive programme falls into one of those categories. No upgrading air defences. Surface-to-air missiles designed against aircraft were not to be given the capability to counter strategic ballistic missiles, and were not to be tested in an ABM mode. Without this rule, either side could have quietly converted a dense air-defence network into a national missile shield. Radar discipline. Early-warning radars were permitted but had to be located on the periphery of national territory and oriented outward. The logic is elegant. A radar in the interior of a country, looking outward, is a battle-management radar for a nationwide defence. A radar on the coast, looking out to sea, gives warning but cannot conduct an interior engagement. Geography was used as a substitute for trust. No transfer. Neither side was to transfer ABM systems or their components to third states or deploy them outside its own territory. 5.4 Verification and the consultative machinery Two features made the treaty durable. The first was #national_technical_means. Each party agreed to verify compliance using its own satellites and sensors, and, crucially, agreed not to interfere with the other side's #national_technical_means and not to use deliberate concealment measures. This was a quiet revolution in international law: reconnaissance satellites, which had been legally ambiguous, were effectively legitimised as instruments of #verification. Without this, the treaty would have needed intrusive inspections that neither side would have accepted in 1972. The second was the #Standing_Consultative_Commission. This was a permanent, confidential body in which each side could raise compliance questions, seek clarification, and negotiate technical understandings. Its value was procedural rather than dramatic. It gave the parties a room to argue in. During periods when the broader relationship collapsed, the commission kept a channel open. Institutional design of this kind is often overlooked by students focused on numbers, but the machinery for handling disputes is frequently what determines whether an agreement survives its first crisis. 5.5 Duration and exit The treaty was of unlimited duration and subject to review every five years. Article XV allowed either party to withdraw on six months' notice if it decided that extraordinary events related to the subject matter of the treaty had jeopardised its supreme interests. This clause, standard in arms control instruments, would later be used exactly as written, and would prove that a legal right can be exercised lawfully and still be strategically contested. 5.6 Agreed statements and the limits of text Alongside the articles themselves, the parties adopted a set of agreed statements, common understandings, and unilateral declarations. These clarified matters such as what counted as an ABM component, how future systems based on new physical principles should be handled, and what would constitute testing in an ABM mode. The provision on future systems is worth pausing over, because it became the centre of the fiercest legal fight of the treaty's life. The parties recognised that entirely new technologies might one day appear, and they agreed that any deployment of such systems would be subject to discussion and agreement. On the ordinary reading, this means that new physics does not create a loophole. On the reading adopted by parts of the United States government in the 1980s, the restriction on development and testing applied only to the technologies existing in 1972, leaving lasers and other exotic concepts free. The dispute demonstrates something students should take seriously. Treaty text is not self-executing. Words such as "component", "develop", and "test" carry enormous strategic weight, and a government determined to find room will usually find it. The defence against this is not better drafting alone. It is a functioning consultative process, a willingness to be embarrassed by one's own negotiators, and domestic institutions willing to hold the executive to the plain meaning of what it signed. 5.7 The treaty at a glance For students, the key features can be summarised as follows. What it banned. A nationwide territorial defence against strategic ballistic missiles. The development, testing, and deployment of sea-based, air-based, space-based, or mobile land-based ABM systems and components. The upgrading of air-defence missiles to strategic interception roles. Interior-facing large radars in the national interior. Transfer of ABM systems to other states. What it allowed. One deployment area per side after 1974, with 100 launchers and 100 interceptors. Fixed, land-based systems at that site. Research, as distinct from development and testing. Early-warning radars on the periphery, facing outward. Defence against threats other than strategic ballistic missiles. How it was policed. National technical means of verification, protected by an obligation not to interfere or conceal. A permanent Standing Consultative Commission. Five-year reviews. How it could end. Six months' notice under Article XV, on a finding that extraordinary events had jeopardised supreme interests. 6. Life Under the Treaty, 1972 to 2001 6.1 A period of reductions The three decades in which the treaty operated were the most productive period in the history of nuclear arms control. SALT II, the INF Treaty, START I, and START II were negotiated in its shadow. Warhead numbers, after peaking in the 1980s, fell dramatically. Correlation is not causation, and it would be careless to credit a single treaty with all of this. But the sequencing is suggestive and the negotiating record is explicit: Soviet and later Russian negotiators repeatedly stated that offensive reductions were possible only because defences were capped. Remove the cap, and each side must retain a larger arsenal as insurance against a future shield. The treaty functioned, in effect, as the foundation on which the offensive reductions were built. 6.2 Compliance disputes The treaty was not free of trouble. The most serious dispute concerned a large phased-array radar constructed by the Soviet Union near Krasnoyarsk in Siberia. It was in the interior of the country rather than on the periphery, which made it, on the plain text of the treaty, a violation. Moscow initially claimed it was for space tracking. The United States rejected the explanation. Eventually, under Mikhail Gorbachev, the Soviet Union acknowledged the problem and the radar was dismantled. The episode is instructive on two levels. It shows that violations occurred. It also shows that the #Standing_Consultative_Commission and sustained diplomatic pressure could produce correction rather than collapse. American conduct was also questioned, particularly regarding large radars and certain test activities. The general pattern was that both sides probed the edges of the text, both sides complained, and the machinery held. 6.3 The Strategic Defense Initiative and the interpretation war The largest stress came in 1983, when President Ronald Reagan announced the #Strategic_Defense_Initiative, an ambitious programme intended to render ballistic missiles, in his phrase, impotent and obsolete. It included concepts for #space_based_interceptors, directed-energy weapons, and layered architectures. This collided directly with the treaty's ban on developing, testing, and deploying space-based ABM systems. The administration responded with a legal manoeuvre known as the broad interpretation, arguing that the ban did not extend to systems based on physical principles that had not existed in 1972. Critics, including several of the original American negotiators, called this a rewriting of the treaty rather than a reading of it. Congress ultimately constrained the executive's freedom to act on the broad interpretation, and the dispute was never fully resolved. Recent archival history has changed how this episode is understood. Bateman (2024) shows, on the basis of declassified American, European, and Soviet documents, that the central controversy surrounding #Strategic_Defense_Initiative was not simply whether the technology could work. It was about the role of military space technology in statecraft, about alliance politics, and about the future of #arms_control in space. The programme did not produce a shield. It did obstruct diplomatic efforts to create new limits in the space domain, and its consequences for the militarisation of orbit are still unfolding. For Soviet leaders, the response to the programme was not reassurance. Gorbachev told Reagan plainly that Moscow had to judge the initiative by its potential capabilities and not by American intentions, and that a shield would be needed only by a country preparing a disarming strike. Reagan's offer to share the technology was not believed. The exchange is a textbook illustration of why the #security_dilemma is so hard to escape: sincerity is not verifiable, and capability is. 6.4 Demarcation and the theatre defence question Through the 1990s, a new problem emerged. Regional or theatre defences, aimed at shorter-range missiles, were not prohibited. But how good could a theatre system get before it became, in effect, a strategic one? The two sides negotiated demarcation understandings, setting thresholds based on interceptor velocity and on the range and speed of target missiles used in testing. These agreements were never fully ratified, but they matter enormously for the present, because the thresholds have since been crossed. Ship-based interceptors now fly faster than the agreed dividing line, and they have been tested against intercontinental-range targets. The technical boundary between regional and homeland defence, which negotiators once tried to draw with numbers, has effectively dissolved (Sankaran, 2025). 7. Withdrawal, 2002 7.1 The decision In December 2001, the United States gave formal notice of withdrawal under Article XV. The withdrawal took effect in June 2002. The United States then proceeded to deploy the Ground-based Midcourse Defense system, or #GMD, with interceptors in Alaska and California. 7.2 The case made for withdrawal The official justification rested on a changed world. The argument had several strands. The Soviet Union no longer existed, and Russia was not an enemy in the Cold War sense, so a treaty designed to manage that rivalry was obsolete. The new danger came from small nuclear programmes and from what officials called #rogue_states, whose leaders might not be reliably deterred by the threat of retaliation. Against a handful of missiles from such a state, a limited defence could be effective. The treaty legally barred the testing needed to develop such defences, and no partial amendment could accommodate the required architecture. American defences would be limited in scale and would not threaten the Russian deterrent, which was far too large to be blunted by a few dozen #interceptors. 7.3 The case made against Critics, including many arms control specialists and a considerable body of academic opinion, answered on four fronts. The deterrence-of-rogues claim was weak. The premise that certain leaders cannot be deterred was asserted rather than demonstrated. Historical cases cut the other way. States with small arsenals have generally behaved with the caution of the weak. The technology did not work well enough. This is the sharpest and most durable objection. Sankaran and Fetter (2022) examined the #GMD record in detail and concluded that after two decades of costly effort the system remained unproven and unreliable, that it had never demonstrated an ability to defeat the simple and cheap #countermeasures a small nuclear state can build, and that its problems stemmed in part from a politically driven rush to deploy in the 1990s. A defence that cannot beat decoys does not solve the #rogue_states problem it was built for. It only unsettles the large adversaries it was supposedly not aimed at. The reassurance was not credible. Telling Moscow and Beijing that a system is limited does not make them plan on that assumption. States plan against capabilities. The United States itself, in the 1960s, refused to be reassured by the modest quality of Soviet defences and instead built its forces against a worst-case estimate. Expecting others to behave differently is not realistic. The precedent was corrosive. Lawful withdrawal is still withdrawal. Removing the foundation stone of a structure invites questions about the rest of it. 7.4 Assessing the arguments The honest verdict is mixed but tilted. The legal case for withdrawal is unassailable. Article XV said what it said, and the notice period was observed. Those who describe the withdrawal as a violation of international law are simply wrong; the treaty contained its own exit. The strategic case is much weaker, and it has aged badly. Two decades later, the #GMD system remains modest in number and unproven against realistic #countermeasures, while the political costs have been substantial. In February 2025, a study by the American Physical Society panel on public affairs concluded that no missile defence system developed to that point had been shown to be effective against realistic #ICBM threats. That is a striking judgement about a programme on which very large sums have been spent. Meanwhile the reaction from Russia and China has been exactly what the treaty's defenders predicted. Whether or not one accepts that withdrawal caused the subsequent unravelling of the arms control system, it is difficult to argue that it helped. 7.5 The withdrawal seen through international law Because students often confuse the legal and the political questions, it is worth separating them cleanly. Under the law of treaties, a state may leave an agreement in accordance with the agreement's own terms. Article XV of the ABM Treaty contained a withdrawal clause of a familiar type, common across arms control instruments and found in the Non-Proliferation Treaty as well. It required notice, a period of six months, and a statement of the extraordinary events that the withdrawing party regarded as having jeopardised its supreme interests. Two features of such clauses are important. First, the judgement about supreme interests is essentially self-assessed. No court or body reviews whether the stated reasons are adequate. Second, the obligation to give reasons is nonetheless real, and it has a political function. It forces the withdrawing state to say something in public that can be examined, remembered, and later tested against events. That is exactly what has happened here. The stated reason was the threat from small nuclear programmes and the need to develop defences against them. Two decades on, that reason can be assessed. The programmes in question have grown. The defences developed against them remain unproven against realistic countermeasures. This does not make the withdrawal unlawful. It does mean that the justification offered has not been vindicated, and the requirement to state a justification is what makes such an assessment possible. There is a further wrinkle that is often overlooked. The treaty had been concluded with the Soviet Union, a state that ceased to exist in 1991. Questions of succession were handled through a 1997 memorandum identifying Russia, Belarus, Kazakhstan, and Ukraine as successors, an instrument that the United States Senate never approved. Some American lawyers argued that the treaty had lapsed with the disappearance of its counterpart. The executive branch did not rely on this argument, preferring the cleaner route of Article XV, but the episode illustrates how the collapse of one of the parties to a bilateral treaty creates legal problems with no tidy answer. 8. After the Treaty: Consequences and Counter-Reactions 8.1 Russia's answer Russia did not respond by building an equivalent shield. It responded asymmetrically, by building offensive systems whose design logic is to make interception irrelevant. The publicly announced family includes a hypersonic glide vehicle that manoeuvres unpredictably in the upper atmosphere instead of following a predictable ballistic arc, a nuclear-powered cruise missile intended to fly around defensive coverage, a long-range nuclear-armed underwater vehicle that avoids the atmosphere entirely, and air-launched and sea-launched #hypersonic_weapons. Whatever the true state of each programme, their advertised purpose has been consistent: to defeat #ballistic_missile_defense. This is the classical response predicted by the offence-defence equation. Defence is answered not by defence but by cheaper offence. As a matter of economics, the defender is on the wrong side of the exchange ratio, because it is nearly always cheaper to add a warhead, a decoy, or a manoeuvring vehicle than to add an interceptor capable of stopping it. 8.2 China's answer China's response has been broader. Its nuclear forces have expanded substantially, its delivery systems have diversified, and it has tested a fractional orbital bombardment concept that would approach from an unexpected direction. American defence assessments have linked these choices in part to long-standing Chinese concern about United States #missile_defense capabilities. The scholarly picture is more subtle than a simple action-reaction story. Chinese debates, examined through Chinese-language sources, reveal a genuine fear that American defences combined with precision conventional strike and cyber capabilities could threaten the survivability of a small retaliatory force (Hiim, Fravel and Troan, 2023). At the same time, other analysts argue that Chinese expansion is driven by political and status considerations that would exist regardless of American programmes. Both drivers can be true at once. What matters for policy is that a substantial part of the expansion is motivated by fears that the #ABM_Treaty was designed to remove. More recently, evidence has emerged of Chinese work on strategic interception of its own, including systems apparently aimed at intercontinental threats and interest in sea-based midcourse capability. If the world's three largest nuclear powers all pursue strategic defences while all also expand offences, the resulting three-way instability will be considerably harder to manage than the two-way version that the 1972 negotiators faced. 8.3 The wider erosion The treaty was the first of the Cold War strategic agreements to fall. It was not the last. The Intermediate-Range Nuclear Forces Treaty ended in 2019 after mutual accusations of violation. The Open Skies Treaty unravelled shortly afterward. Russia suspended its participation in #New_START in 2023, and that treaty expired in February 2026 without a successor, leaving no binding limit on the deployed strategic arsenals of the two largest nuclear powers for the first time since the early 1970s. Careful scholars resist a monocausal story here. Each collapse had its own causes, and the deterioration of the political relationship, culminating in war in Europe, would have strained any agreement. The defensible claim is narrower and still important: the removal of limits on defences removed the anchor of the offensive limits, and once the anchor was gone, the rest of the structure was easier to dismantle. 8.4 Regional defences and alliance politics Between the strategic and the tactical lies a growing category of regional systems. #THAAD batteries have been deployed in several theatres. #Aegis_Ashore sites are operational in Romania and have been established in Poland. Sea-based interceptors are carried on a growing number of ships. These systems create three distinct problems. First, they blur the line the demarcation talks tried to draw. A regional interceptor that can be repositioned, that flies fast enough, and that has been tested against long-range targets is not obviously regional in any meaningful sense. Second, they generate friction with third parties. The deployment of #THAAD in South Korea produced a sustained Chinese reaction concerned less with the interceptors themselves than with the reach of the associated radar. Third, they entangle allies. A defensive deployment on allied soil turns that ally into a participant in a strategic argument it may not have chosen. 8.5 What combat has actually shown The last few years have produced something the ABM debate has always lacked, which is empirical data from real missile attacks rather than scripted tests. The results are sobering for both camps. Defences have performed better than the sceptics of the 1980s expected. Large numbers of short and medium-range missiles have been intercepted in regional conflicts. But interception rates against modern medium-range missiles have been well short of perfect, and defenders have run into two hard limits that no amount of engineering removes: interceptor magazines are finite, and interceptors cost far more than the missiles they destroy. That exchange ratio is the crux. A defence that works but costs ten times more per engagement than the attacker's missile is not a solution to an arms race. It is a way of losing one more expensively. And nothing in these regional engagements says anything reassuring about intercepting an #ICBM warhead travelling several times faster, accompanied by decoys, in the vacuum of space. 8.6 North Korea: the stated reason, examined The clearest test of the post-2002 policy is the case it was built for. The argument was straightforward. A small state might acquire a handful of intercontinental missiles. Deterrence might be unreliable against such a state, either because its leadership is thought to be reckless or because it might use the threat of a strike to deter American intervention in a regional war. A defence sized against a handful of missiles solves the problem without threatening anyone else. Each step in this chain has been challenged. On numbers, the arsenal in question has expanded rather than remained small. A defence sized for a handful of missiles is sized for a threat that no longer exists in that form. On technology, the crucial finding is that the defence has never demonstrated an ability to handle simple countermeasures. Sankaran and Fetter (2022) set out the technical case in detail: a state capable of building an intercontinental missile and a re-entry vehicle is also capable of building balloon decoys, cooled shrouds, or simple manoeuvring, and these are far cheaper than the interceptors sent to defeat them. A midcourse defence that cannot tell a warhead from a balloon in the vacuum of space is not a defence against a competent adversary; it is a defence against an incompetent one. On deterrence, the record does not support the claim that small nuclear states are undeterrable. States that acquire a few weapons have generally behaved as the weak behave: cautiously, and with a keen interest in survival. On the promise of limitation, the trajectory of the programme has broken the promise. Interceptor numbers, sensor reach, and the blurring of the theatre and homeland missions have all pushed in the direction of a bigger system than the original rationale required (Sankaran, 2025). The point of this analysis is not that the threat is imaginary. It is real, and states are entitled to try to protect themselves. The point is narrower and more uncomfortable. The mission that justified leaving the treaty is the mission the resulting system performs least convincingly, while the mission it was promised not to perform is the one it increasingly appears capable of approaching. 9. The Present Debate 9.1 The revival of the shield The idea of comprehensive homeland defence has returned. A large architecture, publicly branded as a national dome, has been directed by the United States government, encompassing layered sensors, ground and sea-based interceptors, and, most significantly, #space_based_interceptors. The space component is the decisive break with 1972. The treaty banned space-based ABM systems explicitly, and it did so because the negotiators understood that a constellation in orbit is inherently a global, nationwide system rather than a point defence. There is no way to build a space layer that defends only against one adversary. 9.2 The case for it Advocates argue that the threat environment has changed beyond recognition. #hypersonic_weapons, cruise missiles, and drones create a spectrum of attack that older systems cannot address. They argue that leaving a homeland undefended is a political choice that no other domain accepts, that technology has advanced, and that costs have fallen with cheaper access to orbit. Some argue that a defence can strengthen deterrence by raising the number of warheads an attacker needs, which complicates attack planning even if the defence is imperfect. 9.3 The case against it Opponents make four arguments. Physics and cost. Space-based interception requires an enormous constellation, because most of the satellites are over the wrong part of the earth at any given moment. Independent scientific assessments have repeatedly concluded that the concept is not practical or cost-effective under realistic conditions. Cost estimates for the current architecture range across an extraordinary span, which is itself a signal that nobody knows what it would take. Countermeasures. The attacker gets a vote. Decoys, manoeuvring warheads, depressed trajectories, and simple saturation are all cheaper than interception. Vulnerability. A defensive constellation in orbit is itself a target, and anti-satellite capability is now widespread. A shield that must be defended is not a shield. Strategic reaction. Most importantly, both Russia and China have stated that such a system is an attempt to acquire a disarming first-strike capability, and both have said they will respond. The response will not be a matching shield. It will be more warheads and more exotic delivery systems. 9.4 Arms control implications Vaddi and Warden (2025) offer the most careful recent treatment of what a large defensive programme means for negotiation. Their conclusion is measured. The #offense_defense_interrelationship will complicate a strategic #arms_control agenda, but does not necessarily destroy it. Russia and China currently retain arsenals large enough that even an ambitious defence would not negate their retaliatory capability, which leaves room, at least in principle, for agreements that limit offensive forces while defences proceed. The counter-consideration is that this room shrinks as arsenals shrink. The lower the offensive numbers, the more a given defence matters. This is the paradox at the centre of nuclear #arms_control: the deeper the cuts, the more sensitive the balance becomes to defences, and the harder it becomes to cut further without addressing them. It is precisely the logic that the 1972 negotiators encoded, and it has not been repealed. 9.5 Emerging technologies Fetter and Sankaran (2025) survey the technologies now pressing on #nuclear_deterrence: artificial intelligence in early warning and command, cyber operations aimed at nuclear systems, autonomous platforms, improved remote sensing that may erode the concealment of submarines and mobile launchers, and #hypersonic_weapons. Their assessment is that the effects are uneven and often exaggerated, but that the direction of travel is toward greater uncertainty about whether retaliatory forces will survive. This matters for the present argument in a specific way. Defences do not need to be perfect to be destabilising. They only need to be good enough to finish off a force that has already been degraded by a counterforce first strike. A defence that could never stop a full retaliatory salvo might still stop a ragged one. That is why the combination of accurate offensive weapons, improved surveillance, and even a modest #ballistic_missile_defense is more worrying than any one of them alone. 10. Discussion: What the Treaty Teaches 10.1 Restraint can be a security policy The first lesson is the hardest for students to accept, because it runs against instinct. Under specific conditions, choosing not to defend yourself makes you safer. The conditions are strict: both sides must have large, survivable retaliatory forces; both must understand the offence-defence equation; and both must be able to verify the other's restraint. Where those conditions hold, mutual vulnerability is not a weakness to be corrected. It is the mechanism that keeps the peace. Where those conditions do not hold, the logic does not apply. This is why the treaty's defenders should not overreach. Defence against a genuinely small threat, in a relationship without mutual deterrence, is a different problem. The mistake made after 2002 was not building any defence. It was building a system whose components, sensors, and testing programme could not credibly be confined to the small-threat mission. 10.2 Institutions matter as much as numbers The second lesson concerns design. The treaty is remembered for the number one hundred. It survived because of #national_technical_means, the ban on concealment, the geographic rules for radars, the prohibition on testing in an ABM mode, and above all the #Standing_Consultative_Commission. Numbers can be argued about. Machinery for arguing about numbers is what keeps disagreement from becoming rupture. Any future agreement, whether between Washington and Moscow, Washington and Beijing, or all three, will need an equivalent body, and building it will be slower and less glamorous than announcing a headline limit. 10.3 You cannot regulate half of a system The third lesson is the #offense_defense_interrelationship itself. An agreement that limits warheads while leaving defences unconstrained asks one side to accept a permanent hedge against its own deterrent. It will not be signed, or it will not last. The current situation illustrates the point negatively. There is no limit on defences and, since February 2026, no limit on offences either. The equation has been freed at both ends. 10.4 Capability speaks louder than intention The fourth lesson is about credibility. Repeated official statements that a system is not directed at a particular country carry little weight, because planners plan against what a system can do, not what its owners say it is for. The only reassurance that has ever worked in this field is a limit that can be verified. This is why #technology_creep is so corrosive. It produces capability without a decision, and therefore capability without a corresponding political commitment to limit it. By the time anyone notices, the shape of the system has already spoken. 10.5 Where the treaty's defenders overstate their case Intellectual honesty requires noting the weaknesses in the pro-treaty position. The claim that the treaty caused thirty years of stability cannot be proven; other factors, including the end of the Cold War itself, were at work. The claim that its end caused the current collapse is likewise unproven; a twenty-year gap separates the two, and the invasion of Ukraine and the deterioration of great-power relations are sufficient explanations on their own. And the treaty did nothing about the arsenals that mattered most. Under its regime, warhead numbers rose to their historical peak. It was a permissive condition for later reductions, not a cause of them. The strongest version of the pro-treaty argument is therefore modest: the treaty made a certain kind of dangerous competition legally and technically difficult, and its absence has made that competition legally and technically easy. That is a real claim, and it is enough. 10.6 Practical implications Five implications follow from the analysis, and they are stated as propositions rather than as recommendations, because the purpose of the article is to equip readers to argue rather than to tell them what to conclude. Any serious negotiation must put defences on the table. The parties will not accept limits on offences while defences are unbounded. This is not stubbornness. It is arithmetic. Attempts to sequence the problem, dealing with warheads first and defences later, have failed repeatedly, and there is no reason to expect a different result now. Transparency is cheaper than reassurance. Public statements of benign intent cost nothing and are worth nothing. Data exchanges, notification of tests, agreed limits on interceptor velocity, and restrictions on testing against long-range targets are all measurable and therefore meaningful. The demarcation talks of the 1990s produced exactly this kind of technical vocabulary, and it exists to be reused. Space is the decisive frontier. The treaty banned space-based interception because that is where the distinction between limited and unlimited defence collapses. Any future architecture that puts interceptors in orbit will be read, correctly, as a nationwide system aimed at everyone. If there is one provision of the 1972 text worth reviving in some form, this is it. Institutions should be built before they are needed. The Standing Consultative Commission was valuable precisely because it existed during the years when nothing else did. A modest, technical, confidential body between rivals is easier to create in calm periods and impossible to create in crises. The three-body problem is genuinely harder. Everything above was designed for two players. With three large nuclear powers, the arithmetic of sufficiency changes: a force adequate against one opponent may be inadequate against two, and cuts that felt safe in a bilateral world may feel dangerous in a trilateral one (Bolt, 2025). The 1972 model cannot simply be copied. It must be rethought. 11. Limitations and Directions for Further Research This study is a synthesis and interpretation, not new empirical research. Its conclusions are only as good as the published record on which it rests, and that record has known gaps. Interceptor performance data is largely classified, and test conditions have often been criticised as unrealistically favourable. Judgements about effectiveness therefore rely on independent scientific panels and on published test histories, both of which have limits. The article treats Russian and Chinese decision-making largely through secondary literature. Work using original-language sources, of the kind now emerging on Chinese nuclear debates, is the most valuable direction for future research. Four questions are worth pursuing. What would a #verification regime for space-based interception even look like, given that satellites are dual-capable and hard to characterise from the ground? Can a trilateral agreement work at all, given that China's arsenal is smaller and its incentives therefore different? Does the empirical record from recent regional missile conflicts change the cost-exchange calculation, and if so, in which direction? Is there a technical definition of limited defence that Russia and China would accept and that the United States could live with? 12. Conclusion The #ABM_Treaty was an unusual document. It asked two frightened, hostile, heavily armed governments to agree that they would remain exposed to each other, and it asked them to write that exposure into law. They did it because they had worked through the arithmetic and concluded that the alternative was worse. For thirty years the arrangement held. It did not end the #arms_race, and it did not prevent arsenals from growing to grotesque size. What it did was fix one half of a two-sided equation, and by doing so it made the other half negotiable. Every offensive reduction treaty of the late twentieth century was built on that foundation. The foundation was removed in 2002 for reasons that seemed sufficient at the time: a vanished enemy, a new class of small threats, a technology that seemed close to working. Two decades later, the enemy has returned in a different form, the small threat has grown rather than been defended against, and the technology remains unable to defeat the cheapest #countermeasures an adversary can field. Meanwhile the responses that the treaty's architects predicted, more warheads and stranger delivery systems, have duly arrived. None of this proves that #missile_defense is always a mistake. It does show that defence and offence cannot be treated as separate policy questions, that reassurance without verification does not reassure, and that a shield which one side calls protection will be read by the other as preparation. The most useful thing a student can take from the 1972 treaty is not nostalgia. It is a method. Ask what the other side must do in response to what you are building. Ask whether your reassurance can be checked. Ask whether the equation you are entering has a stopping point. The negotiators of 1972 asked those questions and did not much like the answers, so they wrote a treaty instead. Whatever comes next, whether a bilateral understanding, a trilateral framework, or nothing at all, will be judged by whether it takes those questions as seriously as they did. #ABM_Treaty #anti_ballistic_missile #strategic_stability #nuclear_deterrence #arms_control #missile_defense #mutual_assured_destruction #security_dilemma #SALT_I #Cold_War_diplomacy #nuclear_disarmament #great_power_competition #international_security #nonproliferation #strategic_studies References American Physical Society Panel on Public Affairs. (2025). Strategic Ballistic Missile Defense: Challenges to Defending the United States Against Nuclear-Armed Intercontinental Ballistic Missiles. American Physical Society. Bateman, A. (2024). Weapons in Space: Technology, Politics, and the Rise and Fall of the Strategic Defense Initiative. MIT Press. Bolt, P. J. (2025). Strategic stability in a new era. Frontiers in Political Science, 6, 1504361. https://doi.org/10.3389/fpos.2024.1504361 Colbourn, S. (2022). Euromissiles: The Nuclear Weapons That Nearly Destroyed NATO. Cornell University Press. Cooper, D. A. (2021). Arms Control for the Third Nuclear Age: Between Disarmament and Armageddon. Georgetown University Press. Fetter, S., & Sankaran, J. (2025). Emerging technologies and challenges to nuclear stability. Journal of Strategic Studies, 48(2), 252-296. https://doi.org/10.1080/01402390.2024.2433766 Hiim, H. S., Fravel, M. T., & Troan, M. L. (2023). The dynamics of an entangled security dilemma: China's changing nuclear posture. International Security, 47(4), 147-187. https://doi.org/10.1162/isec_a_00457 Larsen, J. A., & Smith, S. (Eds.). (2024). Arms Control at a Crossroads: Renewal or Demise? Lynne Rienner Publishers. Narang, V., & Sagan, S. D. (Eds.). (2023). The Fragile Balance of Terror: Deterrence in the New Nuclear Age. Cornell University Press. Sankaran, J. (2025). The motivations and unintended consequences of the US pursuit of missile defense. The Nonproliferation Review. https://doi.org/10.1080/10736700.2025.2484836 Sankaran, J., & Fetter, S. (2022). Defending the United States: Revisiting national missile defense against North Korea. International Security, 46(3), 51-86. https://doi.org/10.1162/isec_a_00426 Vaddi, P. R., & Warden, J. K. (2025). Golden Dome and arms control: Impediment or opportunity? Bulletin of the Atomic Scientists, 81(4), 296-304. https://doi.org/10.1080/00963402.2025.2518872 Zhao, T., & Stefanovich, D. (2023). Missile Defense and the Strategic Relationship among the United States, Russia, and China. American Academy of Arts and Sciences.

The Intermediate-Range Nuclear Forces Treaty, signed in Washington on 8 December 1987 by Ronald Reagan and Mikhail Gorbachev, was the first agreement in history that removed a complete class of nuclear delivery systems from the arsenals of the two superpowers. It banned all United States and Soviet ground-launched ballistic missiles, ground-launched cruise missiles and their launchers with ranges between 500 and 5,500 kilometres, and it required the physical destruction of every declared item within three years. This article offers a structured, student-facing academic review of the treaty across its full life cycle. It traces the strategic and political origins of the agreement in the Euromissile crisis of the late 1970s, examines the negotiation process from the zero option of 1981 to the double zero outcome of 1987, describes the legal architecture and the pioneering verification regime that included continuous portal monitoring at missile production plants, evaluates implementation results including the elimination of 2,692 missiles by mid-1991, and analyses the erosion of the regime between 2014 and 2019 that ended in the withdrawal of the United States on 2 August 2019. The article then assesses the post-treaty environment up to 2026, including the Russian abandonment of its unilateral deployment moratorium in August 2025, the appearance of new ground-launched systems on both sides, and the expiry of New START in February 2026. Using qualitative document analysis and a comparison of realist, institutionalist and domestic-politics explanations, the study argues that the treaty collapsed less because verification failed than because the political conditions that made it valuable disappeared. The conclusion identifies transferable lessons for future arms control: verification technology outlives political will, bilateral treaties are fragile in multipolar systems, and the design of a regime should anticipate technological and geopolitical change rather than freeze a single moment of strategic balance. Keywords: intermediate-range nuclear forces; arms control; verification; deterrence; Euromissile crisis; treaty collapse; European security; nuclear disarmament 1. Introduction Few agreements in the history of #arms_control have carried as much symbolic weight as the Intermediate-Range Nuclear Forces Treaty. Signed in December 1987 and entering into force on 1 June 1988, the #INF_Treaty did something that no previous nuclear accord had attempted. Earlier agreements had limited how many weapons the superpowers could deploy, or where they could test them, or how they could defend against them. The INF Treaty abolished an entire category of weapons. Every United States and Soviet ground-launched ballistic missile and ground-launched cruise missile with a range between 500 and 5,500 kilometres, together with the launchers, support structures and support equipment associated with them, had to be eliminated. The treaty was of unlimited duration. It contained no expiry clause, and for more than three decades it was treated as one of the durable foundations of European security. That foundation no longer exists. On 2 August 2019 the United States completed its withdrawal, citing Russian development and deployment of the #9M729 ground-launched cruise missile, known in NATO reporting as the SSC-8. Russia rejected the accusation, presented counter-accusations of its own, and announced a unilateral moratorium on deploying ground-launched intermediate-range systems in regions where the United States did not deploy them. In August 2025 Moscow declared that this moratorium was over. By 2026 both former parties were fielding or preparing to field the very systems that the treaty had once outlawed, and the last remaining bilateral limit on strategic nuclear forces, #New_START, had expired without a successor. For students of international relations, security studies, diplomatic history and international law, the arc of the INF Treaty is unusually instructive because it contains, in one case, almost every question that the discipline asks about #cooperation between rivals. Why do adversaries who distrust each other agree to give up weapons at all? How can compliance be checked when neither side is willing to open its territory freely? What makes a treaty survive changes in leadership and technology, and what makes it die? Is a legal instrument that constrains only two states useful in a world where a third state, unconstrained, builds thousands of the same missiles? This article addresses those questions through a single extended case study. Its purpose is educational rather than polemical. It is written for advanced undergraduate and postgraduate readers who need a clear, accurate and reasonably comprehensive account of the treaty in one place, with the analytical scaffolding that a journal article normally provides. The argument developed across the following sections is that the #INF_regime should be understood as an unusually successful piece of institutional engineering that was built for a specific and temporary distribution of power. Its #verification system worked. Its elimination provisions were fulfilled ahead of schedule. What failed was not the machinery of the treaty but the political consensus that gave the machinery a purpose. When the strategic environment changed, when new technologies blurred the categories the treaty had defined, and when a rising third power stood outside its scope, the treaty lost the constituency that had sustained it. The article proceeds as follows. Section 2 reviews the scholarly literature and identifies the debates into which this study intervenes. Section 3 sets out the research design and the sources used. Section 4 reconstructs the strategic origins of the crisis, beginning with the Soviet deployment of the SS-20 and the NATO dual-track decision of 1979. Section 5 examines the negotiating process. Section 6 analyses the legal text and its definitions. Section 7 turns to the verification and elimination regime, which was the treaty's most original contribution to the practice of #disarmament. Section 8 reviews implementation and the immediate legacy of the treaty. Section 9 reconstructs the compliance dispute and the collapse. Section 10 compares three theoretical explanations of that collapse. Section 11 describes the post-treaty landscape from 2019 to 2026. Section 12 draws lessons, Section 13 acknowledges limitations, and Section 14 concludes. 2. Literature review The academic literature on the INF Treaty can be divided into four broad clusters, each with distinct assumptions and blind spots. The first cluster is historical. Since the opening of archives in Western Europe, Germany and, more partially, Russia, historians have reconstructed the Euromissile episode in considerable depth. The most complete recent account is Colbourn (2022), who treats the crisis as a story about the internal fragility of the alliance rather than as a simple contest between Washington and Moscow. In her reading, the missiles mattered because they exposed a persistent doubt inside #NATO about whether the United States would really risk its own cities to defend Europe. The edited volume by Gassert, Geiger and Wentker (2021) is the other essential reference. It gathers contributions from historians of the Federal Republic of Germany, of the Soviet Union and of the United States, and it deliberately places the signature of 1987 and the collapse of 2019 within a single analytical frame. Its introductory chapter is unusual in that it treats the rise and the fall of the treaty as parts of one story rather than as separate episodes. Radchenko (2024) and Zubok (2021) supply the wider Soviet context, showing that Gorbachev's willingness to accept asymmetric cuts cannot be explained by Western pressure alone and was bound up with a domestic project of reform and with a redefinition of what Soviet security required. The second cluster is the arms control literature proper, which is concerned with the design and function of agreements rather than with their diplomatic history. Krepon (2021) provides the most comprehensive recent synthesis, arguing that arms control and #deterrence were never opposites but complements, and that the abandonment of the former after the Cold War left the latter dangerously unaccompanied. Within this tradition the INF Treaty is usually treated as the high point of intrusive verification, the moment at which #on_site_inspection moved from a theoretical demand to an operating practice. The literature on verification design, much of it produced by practitioners, emphasises that the treaty's inspection protocol was successful precisely because it was modest in ambition. Inspectors counted, measured and observed. They did not attempt to certify the absence of hidden weapons, a task that remains technically unsolved. The third cluster is the post-collapse policy literature, produced mostly after 2014 and intensively after 2019. It is dominated by two questions: whether the Russian 9M729 really violated the treaty, and what NATO should do in a world without the treaty. Contributions here range from analyses of the European missile balance and the risk of a new #deployment_crisis, to arguments that the treaty had become strategically obsolete because it constrained the United States in the Indo-Pacific while leaving China entirely free. Costlow (2024) develops the most systematic version of the second argument, presenting the treaty as a case study in the unintended consequences of an agreement that solved a European problem while creating an Asian one. Critics of that position reply that the absence of the treaty has not improved the American position in Asia and has reopened a #missile_crisis in Europe that had been closed for a generation. The fourth cluster is the contemporary literature on the general crisis of arms control, in which the INF collapse appears as one node in a longer sequence that also includes the Anti-Ballistic Missile Treaty withdrawal of 2002, the suspension of the Conventional Armed Forces in Europe regime, the withdrawal from the Open Skies Treaty in 2020, the Russian suspension of New START inspections in 2023, and the expiry of New START in February 2026 (Liang, 2026). Lippert (2024) argues that the erosion of conventional arms control in Europe was itself a cause of the war in Ukraine, a claim that, whether or not one accepts it in full, usefully connects the technical world of treaty texts to the political world of war and peace. The annual Nuclear Notebook series and the SIPRI Yearbook provide the empirical backbone for this cluster, tracking arsenals, modernisation programmes and deployments with a level of detail that no single monograph can match (Kristensen, Korda, Johns and Knight, 2025a, 2025b; SIPRI, 2025). Three gaps emerge from this survey. First, historical and policy scholarship rarely meet. Historians explain why the treaty was born but stop in the early 1990s; policy analysts explain why it died but often begin in 2014, treating the earlier history as background. Second, the verification regime is frequently praised but seldom examined closely enough for students to understand exactly what it did and did not permit, which matters because the #compliance dispute of 2014 to 2019 turned on questions that the inspection regime, by then long expired, could no longer answer. Third, the literature is thin on the counterfactual: what would have been required, institutionally, for the treaty to survive. This article does not claim to close those gaps definitively, but it is organised so as to address all three within one continuous narrative. 3. Research design and methodology This study is a qualitative, single-case, longitudinal analysis. The case is the INF regime understood as a whole, from the strategic conditions that produced it in the late 1970s to its aftermath in mid-2026. The choice of a single #case_study is deliberate. Arms control agreements that abolish a class of weapons are rare, effectively unique in the nuclear domain, and the value of the case lies in its depth rather than in its comparability. Three methods are combined. The first is #document_analysis. The primary legal materials are the treaty text itself, its Memorandum of Understanding on Data, the Protocol on Elimination and the Protocol on Inspection. These documents are the authoritative source for the definitions, categories, numbers and procedures reported in Sections 6 and 7. Public statements by governments, alliance communiques and official compliance reports form a second documentary layer, used with the awareness that they are advocacy documents as much as records of fact. The second is #process_tracing. Rather than correlating variables, the analysis follows the sequence of decisions that led from the SS-20 deployment to the dual-track decision, from the dual-track decision to the negotiations, from the negotiations to the double zero outcome, and from the compliance allegations of 2014 to the withdrawal of 2019. Process tracing is appropriate here because the research question is causal but the number of cases is one, and because the observable implications of rival explanations differ mainly in the sequence and timing of events rather than in aggregate outcomes. The third is theory-guided comparison. Section 10 tests three families of explanation against the same evidence: a structural-realist account that emphasises shifting #power_distribution and the rise of China; an institutionalist account that emphasises the erosion of the treaty's own compliance machinery; and a domestic-politics account that emphasises leadership preferences, bureaucratic advocacy and the collapse of the treaty-ratifying consensus in Washington and Moscow. The purpose is not to declare one framework victorious but to show what each captures and what each misses, which is the more honest position for a case with this level of complexity. Two limits on the evidence must be stated at the outset. The intelligence assessments on which the American allegation against the 9M729 rested have never been fully declassified, and the Russian technical rebuttal was never subjected to independent scrutiny. Any account of the compliance dispute is therefore working with an incomplete record, and this article is careful to distinguish between what is documented, what is officially alleged and what is inferred. Second, developments after January 2026 are reported here on the basis of publicly available reporting rather than settled scholarship, and readers should treat the final sections as an interim assessment rather than a historical verdict. 4. Strategic origins: the Euromissile crisis 4.1 The problem of extended deterrence The INF Treaty cannot be understood without the older problem it was designed to manage. From the early 1950s onward, the security of Western Europe rested on an American promise: an attack on a European ally would be treated as an attack on the United States, and would be met, if necessary, with #nuclear_weapons. The promise was credible only as long as European governments believed that Washington would accept the risk of nuclear retaliation against its own territory in order to defend Hamburg or Brussels. As the Soviet Union acquired the capacity to destroy American cities, that belief became harder to sustain. This is the classic dilemma of #extended_deterrence, and every European nuclear debate of the Cold War is in some sense a variation on it. One response was to station American nuclear systems in Europe. Their function was less military than political. Weapons based on European soil, operated by American personnel, were intended to couple the fate of Europe to the fate of the United States so tightly that no Soviet planner could imagine a war confined to the European theatre. The intellectual history of this idea is well documented, and Colbourn (2022) shows that the underlying anxiety about coupling long predates the missiles of the 1980s. 4.2 The SS-20 and the perception of imbalance In 1976 the Soviet Union began replacing its ageing SS-4 and SS-5 missiles with a new system, the #RSD_10 Pioneer, designated the #SS_20 by NATO. The SS-20 was mobile, solid-fuelled, accurate, and carried three independently targetable warheads. It could be moved, hidden and reloaded. Its range covered all of Western Europe from launch sites deep inside Soviet territory, but it could not reach the United States. That last characteristic is the strategic heart of the matter. Because the SS-20 could not strike American soil, it appeared to some European analysts to create precisely the decoupling that Western strategy had been designed to prevent. A weapon that could devastate European targets without touching the American homeland seemed to offer Moscow a rung on the escalation ladder that NATO could not match. Whether this reading was militarily sound was contested at the time and remains contested. The United States had submarine-launched and air-delivered systems in and around Europe, and any nuclear use in Europe would have carried an enormous risk of general war. But perceptions in alliance politics are not decorative. Helmut Schmidt, the West German chancellor, gave the anxiety its most influential public formulation in a London speech in 1977, and once the anxiety was articulated it could not be ignored. 4.3 The dual-track decision of 1979 NATO's answer, adopted in December 1979, was the #dual_track_decision. The first track was military: the United States would deploy 108 Pershing II ballistic missiles and 464 #ground_launched_cruise_missiles in five European countries, namely the Federal Republic of Germany, the United Kingdom, Italy, Belgium and the Netherlands, beginning in 1983. The second track was diplomatic: the United States would simultaneously seek negotiations with the Soviet Union aimed at limiting or eliminating intermediate-range systems on both sides. The two tracks were designed to reinforce each other. Deployment would give Moscow an incentive to negotiate; negotiation would give European governments a political justification for accepting deployment. The #Pershing_II was, from the Soviet standpoint, the more alarming of the two systems. It was fast, accurate and, from bases in West Germany, capable of reaching command targets in the western Soviet Union in roughly ten minutes. The ground-launched cruise missile, by contrast, was slow but small, low-flying and hard to detect. Together they represented a category of threat that Soviet planners had not previously faced on European soil. 4.4 Protest, politics and the domestic dimension The deployment decision produced the largest peacetime protest movement in postwar European history. Hundreds of thousands of people marched in Bonn, London, Amsterdam, Rome and Brussels. The peace movement was not a monolith; it contained pacifists, environmentalists, church groups, communists, conservatives worried about sovereignty and ordinary citizens frightened by the prospect of a nuclear war fought over their heads. Its political effect was uneven. It did not stop deployment, which began on schedule in late 1983. But it made the political cost of deployment extremely high, it destabilised governing coalitions, and it created a permanent domestic incentive for Western leaders to obtain an agreement that would allow the missiles to be removed. This is a crucial and often underappreciated point for students. The #peace_movement did not defeat the missiles, but it changed the value that Western governments placed on a negotiated outcome. Without that pressure, the diplomatic track of 1979 might have remained decorative. Geiger (2021) demonstrates how sharply the domestic constraint operated on the West German government, which had the most to gain from removal and the most to lose from a failed negotiation. 4.5 The nadir of 1983 The years 1982 and 1983 were among the most dangerous of the late Cold War. #Geneva_talks stalled. Soviet negotiators walked out in November 1983 when the first Pershing II missiles arrived in West Germany. Soviet leadership was elderly, ill and deeply suspicious. Western military exercises were interpreted in Moscow, at least by parts of the intelligence apparatus, as possible cover for a real attack. It is against this background that the diplomatic breakthrough of the mid-1980s must be measured. The treaty of 1987 was not the product of a warm relationship. It was the product of a relationship that both sides had come to regard as intolerably dangerous. 5. From the zero option to double zero: the negotiation 5.1 The zero option of 1981 In November 1981 the Reagan administration proposed what became known as the #zero_option: the United States would cancel its planned deployment of Pershing II and ground-launched cruise missiles if the Soviet Union dismantled all of its SS-20, SS-4 and SS-5 missiles. At the time the proposal was widely regarded, including inside the American government, as a negotiating position designed to be rejected. It asked Moscow to destroy hundreds of deployed missiles in exchange for the cancellation of American missiles that did not yet exist in Europe. Soviet negotiators dismissed it accordingly. The historiographical debate about the sincerity of the zero option remains open. Some scholars read it as propaganda aimed at European public opinion. Others, including contributors to the reappraisal volume edited by Gassert, Geiger and Wentker (2021), argue that Reagan personally believed in the abolition of nuclear weapons and that the zero option, whatever its bureaucratic origins, aligned with a genuine presidential conviction. What is not in dispute is that the proposal, once made, could not easily be withdrawn, and that it framed the eventual outcome. 5.2 The walk in the woods In July 1982 the chief negotiators, Paul Nitze for the United States and Yuli Kvitsinsky for the Soviet Union, produced an informal compromise during a private conversation in the Jura hills near Geneva. The so-called #walk_in_the_woods formula would have allowed a limited number of American cruise missiles in Europe in exchange for limits on Soviet SS-20 deployments, and would have cancelled the Pershing II. Both capitals rejected it. The episode is nevertheless instructive: it shows that a negotiated middle position was technically available years before agreement was reached, and that the obstacle was political rather than technical. 5.3 Gorbachev and the reframing of Soviet security The decisive change came with the arrival of Mikhail Gorbachev as General Secretary in March 1985. Gorbachev's approach to security rested on premises that his predecessors had not accepted: that the arms race was economically ruinous, that Soviet security could not be built on the insecurity of others, and that political normalisation with the West was a precondition for domestic reform. Radchenko (2024) and Zubok (2021) both stress that these were not concessions extracted by Western pressure but elements of a strategy whose principal audience was domestic. Gorbachev's first major move on intermediate forces came in 1986, when he offered to eliminate Soviet and American intermediate-range missiles in Europe while leaving aside British and French forces, a linkage that had previously blocked progress. The offer was initially bundled with demands concerning the American #Strategic_Defense_Initiative. 5.4 Reykjavik and the decoupling of the INF question At the #Reykjavik summit of October 1986, #Reagan and #Gorbachev came remarkably close to agreeing on the elimination of far more than intermediate-range systems. The summit ended without agreement because Gorbachev insisted that strategic defences be confined to the laboratory and Reagan refused. Contemporary reporting treated Reykjavik as a failure. In retrospect it was the opposite. The two leaders had discovered how much common ground existed, and in early 1987 Gorbachev detached the intermediate-range question from the dispute over missile defence. That single decision made the treaty possible. 5.5 The double zero The final architecture emerged over 1987. The original zero option covered longer-range intermediate systems, those with ranges between 1,000 and 5,500 kilometres. Once these were to be eliminated, the shorter-range systems between 500 and 1,000 kilometres became more salient, because their retention would have preserved a category of theatre nuclear weapons in Europe. The solution, agreed during 1987, was the #double_zero: elimination of both the longer-range and the shorter-range categories, producing a comprehensive ban across the entire 500 to 5,500 kilometre band. The treaty was signed at the White House on 8 December 1987. The United States Senate gave its advice and consent in May 1988, and the treaty entered into force on 1 June 1988, at the Moscow summit. Contemporary observers understood that the numbers involved were militarily modest compared with the strategic arsenals that remained untouched. Their significance lay elsewhere. For the first time, two nuclear-armed adversaries had agreed to destroy weapons rather than merely to cap them, and to allow each other's inspectors onto their territory to watch it happen. 6. The legal architecture of the treaty 6.1 Scope and core obligations The central obligation of the treaty was stated with unusual clarity. Each party undertook to eliminate all of its intermediate-range and shorter-range missiles, to eliminate the launchers of such missiles, and to eliminate all associated support structures and support equipment, and thereafter not to possess, produce or flight-test any such missiles or produce or possess such launchers. Three features of this formulation deserve attention. First, the ban was categorical rather than numerical. It did not set a ceiling; it set zero. There were no permitted residual holdings, no reserve stockpiles and no conversion options. A #categorical_ban is easier to verify than a numerical limit, because any observed item of the prohibited type is by definition a violation. Second, the ban applied to the missiles and to their launchers, and to the infrastructure that supported them. This mattered because launchers are large, fixed or semi-fixed, and comparatively easy to observe, whereas missiles can be concealed. By outlawing the launcher as well as the missile, the treaty attacked the deployable capability rather than only the object. Third, the ban covered ground-launched systems only. Sea-launched and air-launched missiles of the same range were untouched. This exclusion was not an oversight. It reflected the specific European problem the treaty was designed to solve, the asymmetric verification burden that sea-based systems would have imposed, and the strong American interest in preserving naval flexibility. It also became, three decades later, one of the treaty's principal vulnerabilities, since technological convergence made it increasingly artificial to distinguish a #ground_launched_cruise_missile from an identical missile fired from a ship. 6.2 Definitions and the range bands The treaty defined an intermediate-range missile as one with a range capability between 1,000 and 5,500 kilometres, and a shorter-range missile as one with a range capability between 500 and 1,000 kilometres. #Range_capability, not range as tested or as declared, was the operative concept, and the treaty specified that a missile's range would be determined by the maximum distance it was capable of flying when tested. The definitional work also extended to what counted as a ground-launched cruise missile, to the distinction between a missile and a booster used for other purposes, and to the treatment of research and development items. These definitions did the heaviest legal work in the whole instrument, and they are where the later compliance dispute took place. If a missile is tested at a short range from a ground launcher and at a long range from a different launcher, what is its range capability? If a launch canister is common to two missiles of different ranges, does possession of the canister constitute possession of a prohibited launcher? These questions were not hypothetical in 1987, but they were manageable, because the systems in existence were relatively few and relatively distinct. By 2014 they were neither. 6.3 Categories of covered systems The Memorandum of Understanding attached to the treaty listed the systems covered by name. On the American side these were the Pershing II, the BGM-109G ground-launched cruise missile and the older Pershing IA. On the Soviet side they were the SS-20, the SS-4, the SS-5, the SS-12, the SS-23 and the SSC-X-4 ground-launched cruise missile. The inclusion of the #SS_23, a relatively new short-range system, was a significant Soviet concession and was later criticised inside the Soviet and Russian military establishment as an unnecessary give-away. 6.4 Duration, withdrawal and the absence of an exit ramp The treaty was of unlimited duration. It could be terminated only by withdrawal, and #Article_XV permitted a party to withdraw upon six months' notice if it decided that extraordinary events related to the subject matter of the treaty had jeopardised its supreme interests. This is a standard clause in arms control instruments, and it is worth noting how it functions. It does not require a party to prove that its interests are threatened; it requires only a decision, a notification and a statement of the extraordinary events. In 2019 this clause was the legal instrument of the treaty's death, and no provision of the agreement could have prevented that outcome once the political decision had been taken. 6.5 The Special Verification Commission The treaty established a #Special_Verification_Commission as the forum in which compliance questions would be raised and, ideally, resolved. The commission met regularly in the early years and then fell into disuse as the elimination phase ended. Its revival was attempted during the compliance dispute, but by then the body had neither the technical access nor the political authority to adjudicate a dispute about a missile that neither side was willing to show the other. The fate of the commission is a lesson in institutional maintenance: a dispute-resolution mechanism that is not exercised regularly will not be available when it is finally needed. 7. Verification and elimination: the treaty's most original contribution 7.1 Why verification was the hard problem Arms control between adversaries faces a structural obstacle. Each side has an incentive to promise restraint and then cheat, and each side knows the other has that incentive. Unless promises can be checked, the rational response is to assume the worst, and the agreement collapses before it begins. This is why the history of Cold War arms control is, to a large degree, a history of arguments about #verification rather than about weapons. Before 1987 the standard solution was national technical means, that is, satellite reconnaissance and other remote collection, combined with agreements not to conceal activity from those means. National technical means worked reasonably well for large, fixed objects such as missile silos. They worked badly for mobile, road-transportable systems that could be moved into a forest and covered with a net, which is precisely what the SS-20 was designed to do. 7.2 The inspection regime The INF Treaty solved the problem by inventing, in practice rather than in theory, a layered inspection regime. Its components were as follows. #Baseline_inspections established, at the beginning of the treaty's life, that the data each side had declared were accurate. Inspectors visited declared facilities and confirmed the numbers and types of missiles, launchers and support equipment present. Elimination inspections allowed each side to observe the physical destruction of the other's systems. Destruction was carried out by prescribed methods, including explosive demolition, crushing, cutting and, in the case of some missiles, launching to destruction. The point of prescribing methods was to remove ambiguity: a missile cut into specified pieces cannot be quietly reassembled. Close-out inspections confirmed that a declared facility had ceased the activities associated with the prohibited systems. #Short_notice_inspections permitted each side to demand access to declared sites at limited notice, within a quota, in order to check that prohibited activity had not resumed. #Continuous_portal_monitoring was the most intrusive element and the most novel. American personnel were stationed at the missile final assembly plant at #Votkinsk in the Soviet Union, and Soviet personnel at a corresponding facility at Magna, Utah. Their task was to monitor everything that left the plant, using measurement equipment designed to establish whether a shipping container was large enough to hold a prohibited missile stage. Portal monitoring ran for thirteen years, until 2001. 7.3 What the regime could and could not do The inspection regime was intrusive by the standards of its time, and it was genuinely effective for the task assigned to it. But its limits should be stated plainly, because they explain a great deal about what happened later. The regime verified the elimination of declared items and monitored declared facilities. It did not, and could not, certify that no prohibited missile existed anywhere on the territory of the other party. Inspectors could not roam. They went where the treaty said they could go, when it said they could go there. The system rested on a foundation of declared data, and its logic was to make cheating expensive and detectable at scale rather than to make it impossible. Equally important, the intrusive elements were time-limited. On-site inspection rights ran for thirteen years from entry into force, expiring on 31 May 2001. After that date the treaty's obligations continued indefinitely but its inspection machinery did not. From 2001 onward, compliance had to be assessed by #national_technical_means and intelligence collection alone. When the United States began to suspect, in the early 2010s, that Russia was testing a prohibited cruise missile, there was no inspection right available to test the suspicion, and no functioning mechanism through which Russia could have demonstrated innocence even had it wished to. The treaty had disarmed its own dispute-resolution capacity a decade before the dispute arose. 7.4 Implementation results Implementation was, by any reasonable standard, a success. By the deadline of 1 June 1991, three years after entry into force, both parties had destroyed all declared systems. The totals reported under the treaty were 846 American missiles and 1,846 Soviet missiles, a combined figure of 2,692. The Soviet Union thus destroyed more than twice as many missiles as the United States, an asymmetry that reflected the underlying imbalance the treaty was designed to correct and that later became a grievance in Russian domestic debate. Roughly 850 #inspections were conducted during the inspection period. There were disputes, some of them sharp, over procedures, over access, over the interpretation of dimensions and over the behaviour of monitoring equipment at Votkinsk. None of them escalated into a compliance crisis. The routine, unglamorous management of technical disagreement was itself an achievement, and it is the part of the regime that later analysts most often overlook. 7.5 Multilateralisation after 1991 The dissolution of the #Soviet_Union in December 1991 created a legal problem that the treaty had not anticipated. The obligations of the Soviet Union passed to successor states with relevant facilities or deployments on their territory: Russia, Belarus, Kazakhstan and Ukraine. All participated in implementation, and inspection activity continued on the territory of several of them. This episode is a useful reminder that arms control obligations attach to states in ways that survive the disappearance of the state that signed, and it is also, in retrospect, a small and successful example of the multilateralisation that the treaty's later critics said was impossible. 8. The treaty in its prime, 1988 to 2013 For roughly two decades after entry into force, the INF Treaty enjoyed a status unusual among arms control agreements: it was almost entirely uncontroversial. The missiles were gone. The inspections proceeded and then ended on schedule. European publics stopped marching. The treaty had removed an entire category of weapons from the continent and, with it, a category of political crisis. Three developments during this period nevertheless prepared the ground for what followed. The first was the growing #asymmetry_of_interest between the two parties. For the United States, the treaty cost little, because American power projection relied on sea-launched and air-launched systems that the treaty did not touch. For Russia, the treaty prohibited an entire class of relatively cheap ground-based systems while leaving Russia surrounded by states, none of them party to the treaty, that were free to build them. Russian officials began to voice this complaint publicly in the mid-2000s. In 2007 senior Russian figures raised, in both public and private channels, the possibility of withdrawal, and Russian and American officials jointly proposed at the United Nations that the treaty's obligations be globalised. The proposal went nowhere, because the states that would have had to accept new constraints saw no reason to do so. The second development was technological. The categories on which the treaty rested, ground-launched as opposed to sea-launched, cruise as opposed to ballistic, had been reasonably clean in 1987. By the 2010s, vertical launch systems designed for ships were being adapted for land use, cruise missiles were being built in modular families with common airframes and different fuel loads, and target vehicles used for missile-defence testing had flight profiles that resembled those of intermediate-range weapons. The treaty's definitions did not become wrong. They became contestable, and contestable definitions are the raw material of compliance disputes. The third development was the rise of a third party. #China, which was never a party to the treaty, built a large and modern force of ground-launched ballistic and cruise missiles in exactly the range band the treaty prohibited. Estimates of the proportion of the Chinese inventory that would have violated the treaty vary, but no serious analyst disputes that a large majority of Chinese theatre missiles fell inside the banned band. For American planners concerned with the Western Pacific, this created a strategic asymmetry that the treaty made permanent, since the United States could not respond with ground-launched systems of its own. This is the core of the argument developed by Costlow (2024). Whether the American inability to field ground-launched missiles in Asia was a serious operational handicap, given the availability of sea and air alternatives, is contested. But the argument was politically powerful, and by the second half of the 2010s it had acquired an influential constituency in Washington. 9. Erosion and collapse, 2014 to 2019 9.1 The allegation In July 2014 the United States formally alleged, in its annual #compliance_report, that Russia was in violation of the treaty by producing and flight-testing a ground-launched cruise missile with a range in the prohibited band. The allegation was initially unspecific in public. Over the following years American officials identified the system as the 9M729, known in NATO reporting as the #SSC_8, and stated that Russia had moved from testing to deployment, ultimately identifying several deployed battalions. The American technical case, as publicly presented, was that Russia had tested the same missile from a fixed launcher at a range above 500 kilometres and from a mobile launcher at a range below 500 kilometres, thereby establishing a range capability in the prohibited band while maintaining that the deployed configuration was compliant. Because range capability rather than tested range was the treaty's operative concept, such a pattern would constitute a violation if the facts were as alleged. 9.2 The Russian response Russia denied the allegation. Its public position passed through several stages: first denial that the missile existed as described; then acknowledgement of the 9M729 combined with the assertion that its range was below 500 kilometres; then, in January 2019, a static display of a canister and launcher for foreign attaches and journalists, accompanied by the claim that the missile's range was 480 kilometres. Western governments dismissed the display as insufficient, since a static exhibition of a canister demonstrates nothing about the range capability of the missile inside it. Russia also advanced counter-allegations, of which three were repeated most often. The first concerned the #Aegis_Ashore missile-defence installations in Romania and Poland, which use a land-based version of the #Mk_41 vertical launch system. Because the sea-based Mk-41 can launch Tomahawk cruise missiles, Russia argued that the land-based version constituted a prohibited launcher regardless of what was actually loaded into it. The United States replied that the land-based variant lacked the software and hardware required to fire offensive cruise missiles. The dispute is genuinely difficult, because the treaty prohibited possession of launchers of prohibited missiles, and the question of whether a launcher that could in principle be modified counts as such a launcher was never authoritatively settled. The second counter-allegation concerned target missiles used in American missile-defence tests, which Russia said had characteristics equivalent to intermediate-range ballistic missiles. The third concerned armed unmanned aerial vehicles, which Russia argued met the treaty's definition of a ground-launched cruise missile. Both arguments were rejected by Washington, the first on the ground that the treaty explicitly permitted booster vehicles used for research, the second on the ground that the treaty's definition of a cruise missile presupposed an unmanned, self-propelled vehicle that does not return, which reusable drones are not. 9.3 The collapse The end came quickly. In December 2018 the United States announced that it would suspend its obligations unless Russia returned to compliance within sixty days. NATO allies formally endorsed the American assessment that Russia was in material breach. In February 2019 the United States suspended its obligations and gave notice of #withdrawal. Russia announced a mirror suspension. On 2 August 2019 the withdrawal took legal effect and the treaty ceased to exist. Immediately afterwards, President Vladimir Putin announced a unilateral #moratorium: Russia would not deploy ground-launched intermediate-range systems in any region where the United States did not deploy them, and he invited NATO to reciprocate. NATO governments treated the offer with scepticism, arguing that a moratorium proposed by a state that already possessed the disputed missile would simply freeze an existing Russian advantage. No reciprocal arrangement was agreed, although the offer remained formally on the table for six years. 9.4 Interpreting the collapse Three observations are worth making for analytical purposes. First, the compliance dispute was, in a narrow sense, unresolvable. The treaty's inspection rights had expired in 2001. There was no mechanism by which Russia could have demonstrated the range of the 9M729 to American satisfaction, and no mechanism by which the United States could have demonstrated the incapacity of the Aegis Ashore launchers to Russian satisfaction, short of new and intrusive access that neither government was politically able to grant. A dispute without an adjudication procedure tends to end in the exit clause. Second, the American decision was not driven solely by the Russian violation. Senior American officials stated openly that the treaty's restriction on ground-launched systems in the Indo-Pacific, at a time when China was unconstrained, was itself a reason for withdrawal. This is not a hidden interpretation; it was public policy. The Russian violation supplied the legal and moral justification for a step that had independent strategic motives. Third, European governments were placed in an uncomfortable position and largely chose alliance solidarity over treaty preservation. They accepted the American compliance finding, resisted proposals for immediate counter-deployments, and pressed, without success, for a diplomatic solution. The European failure to convert that discomfort into an initiative is one of the more consequential omissions of the episode. 10. Explaining the collapse: three theoretical readings The narrative in Section 9 describes what happened. It does not explain why. This section sets three explanatory frameworks against the same evidence and assesses what each can and cannot account for. 10.1 The structural-realist reading For #structural_realists, treaties are epiphenomenal. States cooperate when cooperation serves their relative position and defect when it does not. The INF Treaty was concluded in a bipolar world in which the two parties between them held almost all of the relevant capability. By 2019 the distribution of #power had changed. A third state, China, held a large arsenal of the very weapons the treaty banned and was not a party. In such conditions a bilateral prohibition transfers advantage to the state outside it. The predicted outcome, on this account, is exactly what occurred: the party with the greatest incentive to escape the constraint escapes it, and the constraint dies. The strength of this reading is that it explains the timing. It also explains why American withdrawal enjoyed support from officials who were not particularly exercised about the 9M729 as such. Its weakness is that it under-explains the Russian behaviour that provided the occasion. If Russia was the beneficiary of a treaty that constrained American options in Asia, why did Russia risk the treaty by building a prohibited missile? Realist answers exist. Russia may have judged the marginal military value of the 9M729 to exceed the diplomatic cost, or may have expected the violation to remain deniable. But these are auxiliary hypotheses, and they weaken the parsimony that is realism's chief attraction. 10.2 The institutionalist reading #Institutionalists focus on the machinery of cooperation: information, monitoring, dispute resolution and the reduction of transaction costs. On this account the INF regime failed because its institutions atrophied. Inspection rights lapsed in 2001. The Special Verification Commission fell into disuse. No mechanism existed for updating definitions as technology changed. When a serious dispute arose, the regime had no instrument with which to process it, and the parties defaulted to the exit clause because the exit clause was the only functioning provision left. The strength of this reading is that it identifies a genuine and fixable design flaw, and it produces a clear prescription: build treaties with permanent, renewable verification and with standing bodies capable of amending technical annexes. Its weakness is that it may reverse cause and effect. Inspection rights lapsed because the parties, in 1987, saw no need to sustain them, and the commission fell into disuse because nobody used it. Institutional atrophy was itself a symptom of declining political investment, not an independent cause of collapse. Institutions rarely fail when their members want them to work. 10.3 The domestic-politics reading The #domestic_politics reading emphasises the individuals and coalitions that make policy. It notes that the treaty was created by two leaders who wanted it personally and who overrode considerable bureaucratic resistance to obtain it. In the Soviet Union, the general staff resented the asymmetric cuts and the inclusion of the SS-23, and that resentment persisted into the Russian defence establishment. In the United States, a network of officials sceptical of #treaty_based_restraint gained influence in successive administrations, and after 2016 they occupied decision-making positions. The domestic constituency for arms control, which had been broad and bipartisan in the 1980s, had shrunk to a specialist community by the 2010s. On this account, the treaty died because the coalitions that valued it disappeared on both sides. The strength of this reading is that it fits the granular evidence best, particularly the personnel changes and the internal debates documented by Krepon (2021) and by the historical literature. Its weakness is that it risks explaining a structural outcome by contingent means. Had different individuals held office, would the treaty have survived a Chinese arsenal of a thousand intermediate-range missiles and a Russian cruise missile that the United States believed to be prohibited? Possibly, but not obviously. 10.4 An integrated assessment The most defensible position is that the three explanations operate at different levels and are complementary rather than rival. Structural change altered the value of the treaty to its parties. Institutional atrophy removed the means by which the resulting friction could have been managed. Domestic political change determined that the friction would be resolved by exit rather than by renegotiation. Remove any one of the three and the collapse becomes less likely. Structural pressure without institutional decay might have produced an amended treaty; institutional decay without structural pressure might have produced nothing at all, since a dispute mechanism is only needed when there is a dispute. For students, the methodological lesson is as important as the substantive one. Single-level explanations of complex political outcomes are usually elegant and usually incomplete. The task of analysis is not to choose the winning framework but to specify what work each level of explanation performs. 11. The post-INF landscape, 2019 to 2026 11.1 The first years after withdrawal The immediate aftermath of withdrawal was quieter than many had predicted. The United States conducted flight tests of previously prohibited systems within weeks of the treaty's expiry, demonstrating both technical readiness and political intent, but it did not deploy ground-launched intermediate-range missiles to Europe. NATO announced that it would not deploy new land-based nuclear missiles in Europe and would respond in a measured and defensive way. Russia maintained its unilateral moratorium and continued to press, without success, for reciprocal restraint. For a period of roughly three years, the collapse of the treaty produced a legal void rather than an immediate arms race. That relative calm ended with the full-scale Russian invasion of Ukraine in February 2022. The war transformed every element of the European security debate. It destroyed the residual constituency for restraint in most NATO capitals, it accelerated defence spending across the continent, and it placed the question of long-range strike capability at the centre of alliance planning. Lippert (2024) argues that the prior erosion of conventional arms control in Europe was itself among the enabling conditions of the war, an argument that inverts the usual causal story in which war destroys arms control. 11.2 New systems on both sides By the mid-2020s, both former parties were fielding or preparing systems that the treaty would have prohibited. On the American side the principal system is the #Typhon Mid-Range Capability, a ground-based launcher derived from the naval Mk-41 vertical launch system and capable of firing Tomahawk cruise missiles and SM-6 missiles. Typhon has been deployed to the Philippines for exercises and has participated in multinational exercises in Australia. A second system, the Long-Range Hypersonic Weapon, sometimes called #Dark_Eagle, provides a ground-launched hypersonic strike capability. The United States announced in 2024 that it would begin episodic deployments of long-range fires, including Typhon and the hypersonic system, in Germany from 2026, and Germany subsequently indicated an interest in acquiring launchers of its own while European states developed a longer-term programme of their own long-range strike capability. On the Russian side the principal new system is the #Oreshnik intermediate-range ballistic missile, first used against a target in the Ukrainian city of Dnipro in November 2024 and subsequently declared to be in serial production. Russian officials have stated that Oreshnik has been delivered to the armed forces and that deployment to #Belarus was to be completed by the end of 2025. Alongside Oreshnik, Russia has continued to field the 9M729 that occasioned the original dispute, and its short-range Iskander system remains deployed in #Kaliningrad and elsewhere. 11.3 The end of the moratorium On 4 August 2025 the Russian Ministry of Foreign Affairs announced that Moscow no longer considered itself bound by its unilateral moratorium on the deployment of ground-launched intermediate-range and shorter-range missiles. The statement cited the appearance of American systems in Europe and the Asia-Pacific, naming Typhon, the hypersonic weapon and the Precision Strike Missile, whose test at a range beyond its original design maximum Russia treated as evidence of prohibited capability (Arms Control Association, 2025). The significance of this announcement should be understood precisely. It did not change the legal position, because the treaty had already ceased to exist in 2019. What it removed was the last remaining element of declared political restraint. Between 2019 and 2025 the two sides had operated in a legal vacuum tempered by a unilateral commitment. After August 2025 there was neither law nor commitment. 11.4 The wider collapse of the arms control architecture The disappearance of the INF Treaty was not an isolated event but part of a sequence. The United States withdrew from the Anti-Ballistic Missile Treaty in 2002 and from the Open Skies Treaty in 2020. Russia suspended participation in the Conventional Armed Forces in Europe regime and, in February 2023, suspended its participation in New START, ending inspections and data exchanges while stating that it would continue to observe the numerical limits. #New_START expired on 5 February 2026 without a successor and without any negotiation underway (Liang, 2026). Russia had offered in September 2025 to continue observing the treaty's central limits for one year after expiry if the United States reciprocated; the offer was not formally taken up before the expiry date, although both sides indicated that they did not immediately intend to exceed the limits. The result is that, for the first time since the early 1970s, there is no legally binding limit on the strategic nuclear forces of the United States and Russia, and no verified transparency between them. The arsenals themselves remain very large. Open-source estimates put the Russian military stockpile in the region of 4,300 warheads and the American stockpile at a broadly comparable level, with both states engaged in extensive modernisation (Kristensen, Korda, Johns and Knight, 2025a, 2025b; SIPRI, 2025). #China continues to expand its own forces, and the trilateral character of the problem is now the central organising fact of nuclear policy. 11.5 Europe in a post-treaty world For European states the practical questions are uncomfortable and familiar. Should Europe host American ground-launched missiles, accepting the domestic political costs that the 1980s demonstrated so vividly? Should Europe build its own long-range conventional strike systems, which is the direction that current programmes suggest? Should Europe seek to negotiate a new regional restraint regime covering the missiles that now exist on both sides, and if so, with what leverage and what verification? The historical parallel with 1979 is instructive but must be handled with care. The similarities are real: a Russian missile that European publics did not ask for, an American answer that European governments must host, and a domestic debate that no government relishes. But the differences are equally real. There is no mass peace movement of comparable scale. The Russian invasion of Ukraine has changed European threat perceptions in ways that make deployment easier to justify than it was in 1983. And the systems now in question are predominantly conventional rather than nuclear, which changes the political framing even if it does not remove the #escalation_risk that arises when conventional and nuclear systems are indistinguishable in flight (Kristensen, Korda, Johns and Knight-Boyle, 2025). 12. Lessons for the theory and practice of arms control 12.1 Verification technology outlives political will The most durable achievement of the INF Treaty was not the destruction of 2,692 missiles. It was the demonstration that intrusive verification between hostile states is possible. Before 1988, on-site inspection was widely described as unachievable, because no state would permit foreign personnel to stand at the gate of a missile plant and measure what came out. After 1988 that argument was no longer available, and the procedures pioneered under the treaty were carried into START, into cooperative threat reduction and into the design of later regimes. That inheritance survives the treaty's death. Any future agreement, whether between the United States and Russia, or involving China, or covering entirely new technologies, will begin from techniques that were first proved at Votkinsk and Magna. The corollary is sobering. Verification is a solved problem in a way that political will is not. The obstacle to a future agreement is not that we do not know how to count missiles. It is that the parties do not currently wish to be counted. 12.2 Design for change, not for a moment The treaty froze a snapshot of 1987 technology into permanent legal categories. Ground-launched was distinguished from sea-launched; cruise was distinguished from ballistic; a missile's range capability was to be inferred from its flight tests. Those categories were serviceable when systems were few and distinct. They became unstable as launchers converged, as modular missile families proliferated, and as drones and #hypersonic_glide_vehicles emerged that fitted no existing definition cleanly. A treaty that intends to last indefinitely needs a mechanism for revising its technical annexes as technology moves. The Special Verification Commission could in principle have served this purpose, but it was not designed for it and was not used. Future agreements should include a standing technical body with a mandate to update definitions, an obligation to meet whether or not a dispute exists, and a procedure for resolving classification questions before they become accusations. 12.3 Sustain the dispute-resolution machinery The single most consequential structural error in the INF regime was the expiry of inspection rights in 2001. It saved money and political effort in the short term. It meant that when a genuine dispute arose in 2014, the parties possessed no agreed means of establishing facts. Suspicion could neither be confirmed nor dispelled, and in that condition the exit clause becomes the path of least resistance. The lesson generalises: the verification regime should last as long as the obligation. A treaty of unlimited duration should have inspection rights of unlimited duration, exercised at low intensity in quiet periods and available at higher intensity when confidence declines. This is expensive. It is much cheaper than the consequences of collapse. 12.4 Bilateral treaties in a multipolar world The most difficult lesson concerns scope. A bilateral prohibition in a world of three or more relevant powers is inherently unstable, because it transfers advantage to the state outside the agreement and generates a permanent domestic argument for withdrawal in the states inside it. The Russian and American attempt to globalise the treaty in 2007 and 2008 failed, and it is worth being clear about why. States that possessed intermediate-range missiles and had never accepted a constraint on them saw no benefit in accepting one, and the two treaty parties had no leverage to offer beyond a moral appeal. This does not mean that bilateral agreements are worthless. It means that they carry an expiry risk that is a function of the wider distribution of power, and that designers should build in mechanisms for accession, for regional variants and for reciprocal restraint short of full membership. It also means that the current effort to bring a third power into a #trilateral_framework, which has so far been unsuccessful, is not a distraction from arms control but a precondition for it. 12.5 Do not confuse the symbolic and the material Finally, the INF case cautions against two opposite errors. The first is to dismiss arms control as symbolic, on the ground that the missiles eliminated in 1988 to 1991 were a small fraction of the nuclear arsenals then in existence. That is arithmetically true and politically obtuse. The treaty removed a specific crisis from European politics for thirty years, and the crisis returned almost immediately once the treaty was gone. The second error is to treat a treaty as an end in itself, to be preserved regardless of whether it still serves the interests of its parties. Agreements that no longer make sense to those bound by them do not survive on the strength of their historical prestige, and defending them purely on that ground concedes the argument to those who wish to leave. 13. Limitations of this study Three limitations should be stated explicitly. First, the evidential basis of the compliance dispute is asymmetric and incomplete. The American case against the 9M729 depended on classified intelligence that has not been released, and the Russian rebuttal was never made available for independent technical assessment. This article has therefore described the dispute in terms of the claims made by each side and the logic of the treaty's definitions, and has deliberately refrained from adjudicating the factual question of whether the missile's range exceeded 500 kilometres. Readers should understand that this is a limit of the public record and not an evasion. Second, the article relies substantially on English-language and German-language scholarship, and on Russian-language material only as mediated through secondary sources. This introduces a perspective bias that is common in the field and that no single article can fully correct. Third, the closing sections describe events that are still unfolding. The expiry of New START in February 2026, the deployment schedules of new ground-launched systems, and the trajectory of any future negotiation are matters on which the historical record is not yet closed. Conclusions drawn about these developments are provisional. 14. Conclusion The Intermediate-Range Nuclear Forces Treaty was a genuinely remarkable instrument. It eliminated a whole class of nuclear delivery systems, it did so ahead of schedule, and it proved that adversaries could permit each other to inspect their most sensitive military facilities without either state's security collapsing. Judged against what it set out to do, it succeeded. The 2,692 missiles it destroyed did not return, and the crisis that had convulsed European politics for a decade subsided. Judged as an institution intended to last indefinitely, it failed, and the failure repays study. Its verification regime was intrusive but temporary, and it expired long before it was most needed. Its definitions were precise but static, and technology moved past them. Its membership was bilateral in a world that was becoming #multipolar, and the state outside its scope built exactly the arsenal it prohibited. When a compliance dispute finally arose, the regime possessed no functioning mechanism for establishing facts, and the parties defaulted to the exit clause because it was the only clause that still worked. The analysis presented here suggests that no single explanation is sufficient. Structural change made the treaty less valuable to its parties. Institutional decay left them without the means to manage the resulting friction. Domestic political change ensured that the friction was resolved by withdrawal rather than by renegotiation. Each level of explanation contributes something, and the interaction between them is the real story. The situation in 2026 is, by most measures, worse than the situation in 1987. Ground-launched intermediate-range missiles exist again on both sides. The Russian moratorium is gone. New START has expired without a successor, leaving no verified limit on strategic forces for the first time in half a century. Modernisation is proceeding everywhere, and the number of relevant nuclear-armed rivalries has grown. Yet the case also contains grounds for something better than despair. The INF Treaty was negotiated at a moment when the relationship between the parties was, by common consent, at its most dangerous since the Cuban missile crisis. It was concluded not because the two governments trusted each other but because both had concluded that the alternative was intolerable. The techniques they invented to make agreement possible in the absence of trust, the counting rules, the destruction protocols, the portal monitors, the short-notice inspections, still exist and still work. What is missing is not capability but decision. The history of 1987 suggests that such decisions become possible when leaders conclude that the risks of the present course exceed the risks of restraint. Whether that conclusion is reached again, and in time, is a political question, not a technical one, and it is the question that the next generation of scholars and practitioners of #arms_control will have to answer. References Arms Control Association. (2025). Russia cancels intermediate-range missile moratorium. Arms Control Today, September 2025. Colbourn, S. (2022). Euromissiles: The nuclear weapons that nearly destroyed NATO. Ithaca, NY: Cornell University Press. https://doi.org/10.7591/cornell/9781501766022.001.0001 Costlow, M. R. (2024). The INF Treaty, extended deterrence, and assurance: A case study in unintended consequences. Information Series No. 586. Fairfax, VA: National Institute for Public Policy. Futter, A. (2021). The politics of nuclear weapons (new edition). Bristol: Bristol University Press. Gassert, P., Geiger, T., and Wentker, H. (Eds.). (2021). The INF Treaty of 1987: A reappraisal. Goettingen: Vandenhoeck and Ruprecht. https://doi.org/10.13109/9783666352171 Gassert, P., Geiger, T., and Wentker, H. (2021). The reward of a thing well done is to have done it: The rise and fall of the INF Treaty, 1987-2019. An introduction. In P. Gassert, T. Geiger and H. Wentker (Eds.), The INF Treaty of 1987: A reappraisal (pp. 9-40). Goettingen: Vandenhoeck and Ruprecht. https://doi.org/10.13109/9783666352171.9 Geiger, T. (2021). Controversies over the double zero option: The Kohl-Genscher government and the INF Treaty. In P. Gassert, T. Geiger and H. Wentker (Eds.), The INF Treaty of 1987: A reappraisal (pp. 123-153). Goettingen: Vandenhoeck and Ruprecht. https://doi.org/10.13109/9783666352171.123 Krepon, M. (2021). Winning and losing the nuclear peace: The rise, demise, and revival of arms control. Redwood City, CA: Stanford University Press. https://doi.org/10.1515/9781503629615 Kristensen, H. M., Korda, M., Johns, E., and Knight, M. (2025a). Russian nuclear weapons, 2025. Bulletin of the Atomic Scientists, 81(3), 208-237. https://doi.org/10.1080/00963402.2025.2494386 Kristensen, H. M., Korda, M., Johns, E., and Knight, M. (2025b). United States nuclear weapons, 2025. Bulletin of the Atomic Scientists, 81(1), 53-79. https://doi.org/10.1080/00963402.2024.2441624 Kristensen, H. M., Korda, M., Johns, E., and Knight-Boyle, M. (2025). The changing nuclear landscape in Europe. Bulletin of the Atomic Scientists. https://doi.org/10.1080/00963402.2025.2586980 Liang, X. (2026). New START expires as U.S. urges modernized treaty. Arms Control Today, March 2026. Lippert, W. E. (2024). How conventional arms control failures caused the Russo-Ukraine War. Defense and Security Analysis, 40(1), 138-160. Radchenko, S. (2024). To run the world: The Kremlin's Cold War bid for global power. Cambridge: Cambridge University Press. Stockholm International Peace Research Institute. (2025). SIPRI Yearbook 2025: Armaments, disarmament and international security. Oxford: Oxford University Press. Zubok, V. M. (2021). Collapse: The fall of the Soviet Union. New Haven, CT: Yale University Press. #INF_Treaty #Intermediate_Range_Nuclear_Forces #Arms_Control #Nuclear_Disarmament #Cold_War_History #Euromissile_Crisis #Reagan_Gorbachev #NATO_Security #Missile_Verification #Treaty_Collapse #European_Security #Strategic_Stability #Nuclear_Deterrence #Post_INF_World #Security_Studies

This article examines the Treaty between the United States of America and the Russian Federation on Measures for the Further Reduction and Limitation of Strategic Offensive Arms, signed in Prague on 8 April 2010 and known as #New_START. The treaty replaced the expired 1991 Strategic Arms Reduction Treaty and became the last legally binding cap on the two largest nuclear arsenals in the world. Using a qualitative, document based method, the study traces the treaty across four phases: negotiation and entry into force, implementation and compliance, extension in 2021, and breakdown and expiry between 2023 and 2026. It explains the treaty design in plain terms, including the ceiling of 1,550 #deployed_strategic_warheads, the ceiling of 700 deployed delivery vehicles, and the ceiling of 800 deployed and non deployed launchers. It also explains the machinery that made those numbers meaningful, such as #on_site_inspections, twice yearly #data_exchanges, missile #unique_identifiers, and limited #telemetry_exchange. The article argues that New START succeeded as an accounting and #transparency instrument while failing as a political instrument, because it never adapted to three shocks: the collapse of the wider treaty architecture, the war in Ukraine, and the rapid growth of the Chinese nuclear arsenal. The treaty expired on 5 February 2026 without a successor, leaving both parties bound only by informal restraint. The study concludes that the value of New START lay less in the warhead numbers themselves than in the routine, boring, verifiable knowledge it produced, and that any future framework must recover this function even if it cannot recover the treaty form. Keywords: New START; nuclear arms control; strategic stability; verification; deterrence; United States and Russia; nuclear nonproliferation; treaty expiry. 1. Introduction For most of the nuclear age, the United States and the Soviet Union, and later the United States and the Russian Federation, agreed on one unusual thing. They agreed to tell each other, in careful and checkable detail, how many weapons they had aimed at each other. This was not friendship. It was self interest. Each side judged that knowing the size and shape of the other side's forces was safer than guessing. New START was the last treaty built on that judgement. It was signed in Prague on 8 April 2010 by President Barack Obama and President Dmitry Medvedev, it entered into force on 5 February 2011, and it expired on 5 February 2026. Across fifteen years it set the outer boundary of the strategic nuclear competition. It replaced #START_I, which had lapsed in December 2009, and it absorbed the function of the 2002 Strategic Offensive Reductions Treaty, sometimes called #SORT, which was due to end in 2012 and which had almost no verification content of its own. The importance of the treaty was not simply that it lowered numbers. It did lower them: the ceiling of 1,550 deployed warheads was roughly seventy four percent below the 6,000 warhead level counted under START I and about thirty percent below the upper SORT ceiling of 2,200. But the deeper importance was structural. New START gave the two states a shared language for counting #strategic_offensive_arms, a shared set of rules for what counted and what did not, and a shared right to send inspectors to look. It converted an argument about intentions into an argument about data. That conversion is the core product of #nuclear_arms_control. The end of the treaty therefore matters more than the arithmetic suggests. As of the writing of this article in mid 2026, the United States and Russia are for the first time since 1972 without any legally binding limit on their strategic nuclear forces. Russia has said it will keep observing the old ceilings so long as Washington does the same. Washington has said it wants a newer and broader agreement that also captures China. Neither statement is a treaty. Neither statement can be verified. The world has moved from a regime of checked restraint to a regime of declared restraint, and the difference between those two things is exactly the difference that New START was designed to protect. This article has four aims. First, it explains what the treaty actually contained, in language that a student meeting the subject for the first time can follow. Second, it reconstructs the political history of the treaty, from the 2009 negotiations through the 2021 extension and the 2023 Russian #treaty_suspension to the expiry of 2026. Third, it assesses the treaty against the standards that arms control theory sets for itself, namely #crisis_stability, #arms_race_stability, and cost reduction. Fourth, it considers what comes next, including the two peer problem, the possibility of warhead uploading, and the design questions that any successor arrangement must answer. The argument advanced here is that New START was a well built machine placed in a room that was slowly being demolished around it. Its technical design was strong and, on the evidence, both parties stayed within its numerical ceilings even after the political relationship collapsed. Its weakness was that it addressed only one slice of the nuclear problem, a slice defined in 1991 and only lightly updated in 2010, while the rest of the problem grew larger and stranger. Understanding that mismatch is the most useful thing a student of this treaty can take away. 2. Literature Review and Theoretical Framework 2.1 The classical theory of arms control Modern arms control theory grew out of work published in the United States in the early 1960s, and its central claim has proved durable. The claim is that adversaries can have shared interests even while remaining adversaries, and that the most important shared interest is avoiding a war that neither wants. From this claim three goals follow. The first goal is crisis stability. This means that in a moment of extreme tension, neither side should feel pressure to strike first. Pressure to strike first appears when a leader believes that waiting will cost them their own forces. Arms control tries to remove that belief by making retaliatory forces survivable, for example by favouring submarines at sea over vulnerable missiles in fixed silos. The second goal is arms race stability. This means that neither side should feel forced to build more simply because the other side might be building more. Uncertainty drives worst case planning, and worst case planning drives budgets. transparency reduces uncertainty, and reduced uncertainty reduces the political demand for expansion. The third goal is cost reduction. Nuclear forces are extremely expensive. If two rivals can hold each other at risk with fewer weapons, both can spend less without losing security. New START can be read as a straightforward application of these three goals. It permitted survivable forces, it constrained rapid growth, and it capped the size of the deployed force that had to be paid for. The literature of the last five years, however, has increasingly asked whether this classical framework still fits the world it is applied to. 2.2 Contemporary debates Recent scholarship has moved in several directions relevant to this study. One body of work argues that #deterrence has entered a genuinely new age, in which the simple two player logic of the Cold War no longer holds. Narang and Sagan and their contributors describe a fragile balance in which regional nuclear powers, new technologies, and asymmetric doctrines create escalation paths that classical theory did not anticipate. If the balance is fragile in this way, a treaty that counts only intercontinental range systems is addressing a narrowing share of the danger. A second body of work focuses on entanglement between nuclear and non nuclear systems. Command, control, and early warning networks are increasingly shared between conventional and nuclear missions, and they are increasingly digital. An attack intended as conventional may be read as the opening of a nuclear attack. New START protected #national_technical_means, meaning satellites and other remote sensors, from interference, and that protection quietly supported the early warning systems on which crisis decision making depends. This protection is rarely discussed but it is one of the treaty's most underappreciated contributions. A third body of work examines the collapse of the wider architecture. Between 2002 and 2023 the United States withdrew from the #ABM_Treaty, both sides left the #INF_Treaty, and the #Open_Skies_Treaty unravelled. Analysts writing after 2022 have described a compounding effect: each lost agreement removed a channel of contact, and the loss of channels made the remaining agreements harder to defend politically. A fourth body of work, growing quickly since 2023, concerns the two peer problem. The United States now faces two states with large or growing arsenals rather than one. This has produced a genuine analytical disagreement. Some argue that bilateral ceilings designed for a one peer world are now dangerous for the United States, because they cap American forces while leaving #China_nuclear_buildup untouched. Others argue that abandoning ceilings simply invites everyone to build, and that no amount of American expansion can restore the simple certainty of a two player game. 2.3 The gap this article addresses Most existing writing on New START falls into two groups: technical fact sheets that explain the counting rules without political context, and policy commentary that debates the politics without explaining the counting rules. Students need both. This article therefore attempts an integrated account: the mechanics and the politics, read together, across the full life of the treaty from signature to expiry. 3. Methodology This study uses qualitative documentary analysis. It is a form of research well suited to treaty studies, because treaties are, in the end, texts that produce behaviour, and both the texts and the behaviour leave a public record. Three categories of material were used. The first category is primary treaty material and official statements. This includes the structure of the treaty itself, which consists of a preamble and sixteen articles, a protocol containing definitions and verification procedures, and technical annexes to that protocol. It also includes official aggregate data releases, government fact sheets, and formal statements made by both governments before and after expiry. The second category is expert institutional analysis. This includes assessments published between 2021 and 2026 by research organisations that specialise in nuclear questions, along with the annual open source estimates of world nuclear forces produced by Swedish and American research groups. These estimates are used with care, because they are estimates. Neither state publishes a full inventory, and after the collapse of data exchanges in 2023 the uncertainty in these figures increased. The third category is peer reviewed scholarship on deterrence, #strategic_stability, and #verification_regime design, drawn mainly from work published in the last five years. The analytical approach is process tracing. The article follows the treaty through identifiable phases and asks, at each phase, what changed, why it changed, and what effect the change had on the treaty's ability to do its job. Judgement is applied where the evidence is contested, and where it is contested the article says so rather than pretending to a certainty the record does not support. Three limitations should be stated at the outset. First, the classified record is not available, so conclusions about intentions are inferences. Second, events after February 2026 are recent, and early interpretations of recent events are often revised. Third, the study relies mainly on English language sources, which may under represent internal Russian debate. These limitations are revisited in Section 12. 4. Historical Background: The Road to Prague 4.1 From limitation to reduction Nuclear arms control began as an attempt to slow growth rather than reverse it. The Strategic Arms Limitation Talks, known as #SALT, opened in November 1969 and produced an interim agreement in 1972 alongside the ABM Treaty. The logic was elegant. If both sides agreed not to build large defences against ballistic missiles, then neither side would need to build ever more offensive missiles to overwhelm those defences. Limiting defence was therefore a way of limiting offence. A second agreement, SALT II, was signed in 1979 but never ratified by the United States Senate. Yet both states chose to observe its terms informally for several years. This episode matters for the present day, because it established a precedent: parties can respect the substance of an agreement that has no binding force. That precedent has been invoked repeatedly in 2025 and 2026. The end of the Cold War changed the ambition from limitation to reduction. START I, signed in July 1991, was the largest arms reduction agreement in history. It capped each side at 6,000 accountable warheads on 1,600 delivery vehicles and, crucially, it built an intrusive verification system with on site inspection, continuous monitoring at a missile production facility, and detailed notifications. Much of the machinery later inherited by New START was invented here. START II, signed in 1993, would have banned land based missiles carrying multiple independently targetable warheads, a category usually shortened to #MIRV. It never entered into force. A third agreement, START III, was discussed but never concluded. 4.2 The Moscow Treaty and the verification gap In 2002 the United States and Russia signed the Strategic Offensive Reductions Treaty, or SORT, sometimes called the Moscow Treaty. It was famously short, only a few pages, and it committed each side to reduce operationally deployed strategic warheads to between 1,700 and 2,200 by the end of 2012. It contained no counting rules of its own and no verification regime. It relied on the machinery of START I to give its numbers meaning. This created a trap. START I was due to expire in December 2009. When it expired, SORT would be left as a promise with no way to check it. The two states would still have obligations, but no inspectors, no notifications, and no agreed definitions. This is precisely the problem that New START was created to solve, and it is precisely the problem that has returned in 2026. 4.3 The negotiation, 2009 to 2010 Talks opened in 2009 in a period of relative warmth in the relationship, often described at the time as a #reset_policy. The negotiating teams worked under real time pressure, because START I expired in December 2009 and the negotiators wanted to close the resulting gap quickly. The lead American negotiator later published a detailed account of the process, and her narrative underlines how much of the difficulty lay not in the headline numbers but in definitions, conversion procedures, and the treatment of telemetry exchange. Three disputes shaped the final text. The first was #missile_defence. Russia wanted binding limits on American defensive systems, arguing that a large defence would undermine the value of its retaliatory force. The United States refused any binding limit. The compromise was a preambular acknowledgement that offensive and defensive arms are interrelated, plus a narrow prohibition on converting offensive missile launchers into missile defence launchers. Russia attached a unilateral statement reserving the right to withdraw if American defences grew in a way it judged threatening. This unresolved argument would resurface in 2025 and 2026. The second was telemetry exchange, meaning the flight test data that missiles transmit during launch. The United States valued it; Russia was reluctant. The result was a limited and reciprocal exchange rather than the fuller exchange found in START I. The third was the treatment of #heavy_bombers, discussed in Section 6. The treaty was signed on 8 April 2010 in Prague, which is why it is sometimes called the #Prague_Treaty. #ratification was not automatic. The #US_Senate gave its advice and consent on 22 December 2010 by a vote of seventy one to twenty six, after the administration committed to a substantial long term investment in the American nuclear weapons complex. That commitment is itself historically important: the political price of the last major reduction treaty was the beginning of the modernisation programme that now shapes American strategic forces. The Russian #State_Duma and Federation Council completed their own procedures in January 2011, and the treaty entered into force on 5 February 2011. 5. The Architecture of the Treaty 5.1 The three central limits The heart of New START is a set of three ceilings, known collectively as the #central_limits. Each party had seven years from entry into force to reach them, meaning a deadline of 5 February 2018. Both parties met the deadline and both remained at or below the ceilings while the treaty was in force. The first ceiling is 1,550 accountable deployed strategic warheads. The second ceiling is 700 deployed delivery vehicles, meaning deployed intercontinental ballistic missiles, or #ICBM, deployed submarine launched ballistic missiles, or #SLBM, and deployed heavy bombers equipped for nuclear armaments. The third ceiling is 800 deployed and non deployed #launchers, meaning ICBM launchers, SLBM launchers, and heavy bombers equipped for nuclear armaments. This third ceiling is easy to overlook but it does important work. By capping empty launch tubes and silos as well as loaded ones, it limits the physical capacity available for a rapid #breakout, which is the term for a sudden expansion beyond agreed levels. Within these ceilings each party had complete freedom to structure its forces as it wished. This flexibility was a deliberate design choice, and it is one reason the treaty survived changes of government in both capitals. 5.2 The counting rules and their consequences Counting rules decide what the numbers mean, and the counting rules of New START have two features that a serious student must understand. The first feature is that warheads on ballistic missiles are counted by actual loading. Under START I, a missile type was attributed a fixed number of warheads regardless of what it actually carried. Under New START, inspectors count the actual re entry vehicles on a randomly selected deployed missile. This is a more honest measure of the deployed force, and it is one of the treaty's genuine improvements. The second feature is the #bomber_counting_rule. Each deployed heavy bomber counts as one warhead, even though a bomber can physically carry many more, in the range of sixteen to twenty depending on the aircraft. This is an inherited fiction. It was carried over from START I on the reasoning that bombers are slow, recallable, and unsuitable for a surprise first strike, so they contribute less to instability than missiles do. The practical effect, however, is that the true number of warheads that could be delivered by treaty accountable forces was always somewhat higher than 1,550, by perhaps a few hundred. The honest way to describe New START is therefore this: it was a strict cap on missile warheads, a loose cap on bomber warheads, and no cap at all on warheads in storage. 5.3 What the treaty did not cover The list of exclusions is long, and it explains much of the later political trouble. The treaty did not limit #non_strategic_nuclear_weapons, meaning shorter range systems. Russia is generally assessed to hold considerably more of these than the United States, and their exclusion became a central American complaint. The treaty did not limit warheads in reserve. Both states hold a large #hedge_stockpile of non deployed warheads that can, given time, be loaded onto existing missiles. This is the single most important gap, because it means the deployed ceiling constrains posture rather than capability. The treaty did not limit missile defence, beyond the narrow conversion ban already described. The treaty did not clearly cover several new Russian systems announced after 2018. The #Avangard #hypersonic_glide_vehicles, being carried on an intercontinental ballistic missile, were accepted as accountable, and the #Sarmat heavy missile was likewise treated as accountable once deployed. But the #Poseidon nuclear powered underwater vehicle and the #Burevestnik nuclear powered cruise missile fall outside the treaty's definitions, since they are neither ICBMs nor SLBMs nor heavy bombers in the treaty sense. Whether such systems should be accountable became an open dispute and remains one. The treaty did not cover any third country. It was strictly bilateral. In 2010 this seemed obvious and reasonable. By 2026 it was the central objection raised against renewing it. 6. The Verification Regime If the ceilings are the skeleton of New START, the verification regime is its nervous system. Students often skip this part because it is technical. That is a mistake, because the verification machinery is the part that actually died first, and its death is what killed the treaty. 6.1 Inspections The treaty allowed each party up to eighteen on site inspections per year, divided into two types. Type One inspections took place at bases where deployed systems are located. They served two purposes at once: confirming the declared number of delivery vehicles at the site, and confirming the number of warheads actually loaded on a randomly chosen deployed missile. The inspected party had to designate a missile and allow the inspectors to count its re entry vehicles, subject to procedures that protect sensitive design information. Ten such inspections were permitted each year. Type Two inspections, of which eight were permitted each year, covered facilities holding non deployed items, conversion and elimination sites, and storage facilities, and confirmed that items declared as converted or eliminated really were. 6.2 Data exchange and notifications Twice a year, on 1 March and 1 September, the parties exchanged a full database covering every treaty accountable item: its type, its location, and its status. Between these exchanges, a continuous stream of notifications reported movements, new items, eliminations, and changes of status. Over the treaty's life the parties exchanged many thousands of such notifications. The value of this stream is easy to underestimate. It is not dramatic. It is a slow, dull, cumulative picture of what the other side is doing, and it is precisely that dullness that allows analysts and intelligence services to notice when something unusual happens. 6.3 Unique identifiers Every accountable missile and every heavy bomber carried a unique identifiers marking. This is close to a serial number. It allowed each item to be tracked over time and made it far harder to hide a stock of extra missiles, since an inspector could ask where a specific item had gone. 6.4 National technical means Both parties committed not to interfere with the other's national technical means of verification, meaning reconnaissance satellites and similar remote systems, and not to use concealment measures that would defeat them. This obligation had a quiet second effect. Many of the same space based assets that watch missile fields also provide early warning of a launch. A commitment not to attack or blind them is, indirectly, a commitment to preserve the other side's ability to see clearly in a crisis. 6.5 The Bilateral Consultative Commission The treaty created a standing body, the #Bilateral_Consultative_Commission, which met at least twice a year in Geneva. It was where implementation questions were raised and, usually, quietly resolved. Russia raised objections to American procedures for converting Trident launch tubes and B-52 bombers to non nuclear status, arguing that the conversions could be reversed. The United States defended its procedures. The dispute was aired, examined, and managed. That is what a #compliance body is for, and its absence is now keenly felt. 6.6 What the regime achieved The verification system worked. From 2011 to early 2020 inspections were conducted regularly, data flowed, and both governments repeatedly certified the other side's compliance with the central limits. The United States determined annually, across administrations of both parties, that Russia was meeting its treaty obligations. This is an unusual record and it deserves to be stated plainly, because it complicates the common assumption that Russia never keeps agreements. On the numbers, for most of the treaty's life, it did. 7. Implementation, 2011 to 2020 7.1 Meeting the limits Both states reached the central limits by the February 2018 deadline, and the way each reached them is instructive, because it reveals the different logic of the two force structures. The United States relies heavily on submarines. It reduced its deployed force in part by removing four of the twenty four launch tubes on each Ohio class ballistic missile submarine from nuclear service, and by converting a portion of its B-52 bomber fleet to conventional only roles. These conversions were designed to be, in the American view, effectively permanent. Russia disagreed, and this disagreement was raised repeatedly in the Bilateral Consultative Commission. Russia relies more heavily on land based missiles, many of them carrying multiple warheads. It reduced its deployed force partly through the retirement of old Soviet era systems and partly through the loading choices it made on newer missiles. Because so many Russian missiles are MIRV capable but not fully loaded, Russia retained a substantial technical ability to add warheads quickly if the treaty were to end. This asymmetry, sometimes called #upload_capacity, sits at the centre of every serious debate about what happens after the treaty. It is worth being precise here, because the point is often stated carelessly. Both sides have upload potential. The United States has it too, through unfilled Trident tubes, reconvertible launchers, and a large reserve of stored warheads. The difference is one of speed and scale, and analysts disagree about how large the difference actually is. What is not in dispute is that the deployed ceiling was never a ceiling on capability. It was a ceiling on posture. 7.2 The final years of full implementation By 2019 the wider architecture was already crumbling. The INF Treaty ended in 2019 after American accusations of Russian violation and Russian counter accusations. The Open Skies Treaty unravelled in 2020 and 2021. New START survived these events, and by the end of the decade it was the only major agreement left standing. Then the machinery stopped for a reason that had nothing to do with politics. In March 2020 both parties suspended on site inspections because of the COVID-19 pandemic. This was a mutual, sensible, temporary decision. It never became untemporary in the way anyone intended. Inspections did not resume in 2021, and they did not resume in 2022, and by the time the political will to resume them might have returned, the political relationship had collapsed. The pause of 2020 turned out to be the beginning of the end, and it is a reminder that institutions can die of neglect as easily as of attack. 8. The 2021 Extension The treaty had a ten year term with a provision for a single extension of up to five years. There was no legal route to a second extension. This design reflected an assumption held by the negotiators in 2010: that a successor agreement would be negotiated in the intervening years, as had happened repeatedly in the past. The assumption failed. Between 2017 and 2020 the American position on extension was uncertain. Late in 2020 Washington proposed a shorter extension conditioned on a freeze covering all warheads, including the reserve stockpile, and on improved verification measures. Moscow refused the conditions. The matter was resolved quickly after the change of American administration in January 2021. Within days of the inauguration the two presidents agreed by telephone to a full five year extension. The Russian parliament moved rapidly, and the extension was formalised in early February 2021, only days before the treaty would have lapsed. The new expiry date was set for early February 2026. The extension bought five years. It did not buy a plan. Officials on both sides spoke of using the time to negotiate a successor covering all warheads and possibly other systems, and a #strategic_dialogue was launched in mid 2021 for that purpose. It produced procedural progress and little else. The dialogue was suspended after the full scale invasion of Ukraine in February 2022 and never meaningfully resumed. The five years were spent, and at the end of them there was no successor text, no negotiating framework, and no agreed agenda. 9. Breakdown: Suspension and Countermeasures, 2022 to 2023 9.1 The war and the treaty The Russian invasion of Ukraine in February 2022 damaged New START in three ways at once. It destroyed the political constituency for cooperation in Washington. It gave Moscow a reason to treat treaty implementation as a bargaining chip rather than a shared interest. And it introduced explicit #nuclear_signalling into an ongoing conventional war, which raised the perceived stakes of every technical dispute. Through 2022 the parties tried and failed to restart inspections. A session of the consultative commission scheduled for late 2022 was postponed by Russia at short notice. Moscow argued that sanctions and travel restrictions made reciprocal inspection impossible in practice, and that Washington could not expect the benefits of inspection while supporting Ukraine militarily. Washington replied that the treaty served the interest of both parties regardless of other disputes, and that the two questions should be kept separate. 9.2 The suspension of February 2023 In early 2023 the United States formally determined that Russia was not complying with the treaty, on the grounds that it was refusing inspections and refusing to convene the consultative body. On 21 February 2023 President Putin announced that Russia was suspending its participation in New START. The word suspension has no meaning in the treaty text. There is no provision for suspension. There is a provision for withdrawal on three months notice, and Russia did not use it. What Russia did was to stop implementing the verification and notification provisions while stating that it would continue to observe the numerical central limits. It was, in effect, an attempt to keep the parts of the treaty it valued and discard the parts it did not. This is why the treaty suspension is usually described in Western legal analysis as a violation rather than a lawful act. The practical consequences were immediate. Inspections were dead. The twice yearly data exchanges stopped. Russia ceased providing the notifications that had been flowing for twelve years. Russia did, however, state that it would continue to provide ballistic missile #launch_notifications under a separate 1988 agreement, and that it would continue to observe the deployed ceilings. 9.3 American countermeasures The United States responded with what it described as lawful and reversible #countermeasures, taking effect from June 2023. It stopped providing Russia with day to day treaty notifications on the status and location of accountable items. It stopped providing telemetry data. It declined to accept Russian inspectors on American territory, and revoked the visas of the inspection teams. It withheld its own data declarations from Russia, while continuing for a period to publish aggregate figures publicly. Washington also stated that it would continue to observe the central limits so long as Russia did the same, and that it wished to see full implementation restored. The result was a strange half life. From 2023 onward, New START existed as a legal instrument that neither party was fully implementing and both parties claimed to respect. Analysts began describing it as a zombie treaty: formally alive, functionally dead. Both governments continued to assert that they were within the ceilings, and open source estimates broadly supported that assertion, but nobody could check it in the way the treaty had been designed to allow. The regime had shifted from verified restraint to asserted restraint, and this shift, more than any change in warhead numbers, is what analysts now identify as the decisive loss. 10. The Endgame, 2025 to 2026 10.1 The Russian proposal In September 2025 President Putin made a public proposal. Russia would continue to observe the quantitative central limits of New START for one year after the treaty's expiry, until early February 2027, if the United States would do the same. The offer carried conditions. Russia expected reciprocity and expected Washington to refrain from steps that Moscow judged destabilising, with missile defence expansion specifically named. Notably, the offer did not include restoring the verification regime. It was an offer of parallel political commitments, not of checkable obligations. This was, in its structure, an echo of the informal observance of the unratified SALT II agreement in the early 1980s. Precedent existed. Whether the precedent was reassuring or merely nostalgic was disputed. 10.2 The American debate The American response was slow and, in the end, absent. The internal debate turned on three questions. The first question was the #two_peer_problem. American officials and analysts argued that the ceilings had been designed when the United States faced one nuclear peer. With China nuclear buildup proceeding rapidly, from roughly two hundred fifty warheads a decade ago to an estimated six hundred operational warheads now, with official American projections of around one thousand by 2030, the same ceiling that once provided sufficiency might no longer do so. On this view, accepting the old limits would be accepting a self imposed disadvantage. The second question was compliance and confidence. Without inspections, the United States could not confirm Russian adherence. A commitment that cannot be checked, critics argued, provides the political appearance of restraint without its substance, and might even conceal a Russian advantage in #reconstitution. The third question was leverage. Some argued that a one year observance would buy time for serious negotiation, would steady allies, and would allow both states to appear responsible at the 2026 review conference of the #NPT. Others argued that agreeing to the old limits would remove the pressure Moscow needed to feel before it would negotiate on the things Washington actually cared about, namely non strategic nuclear weapons and total warhead numbers. Supporters of accepting the offer, including the American official who had led the original negotiation, warned publicly that if Washington abandoned the ceilings while lacking the capacity to expand quickly, Russia might simply upload faster and gain a temporary numerical advantage. This is a serious argument, and it does not depend on trusting Moscow. It depends only on comparing the speed at which each side can add warheads to missiles it already possesses. 10.3 Expiry The treaty expired on 5 February 2026. There was no formal American response to the Russian proposal. There were reports of late, informal contacts on the margins of other negotiations, and reports that an understanding to observe the spirit of the agreement might be reached, but nothing was signed, and officials made clear that a legal extension was not available in any case, because the treaty permitted only the single extension already used in 2021. On the day of expiry the American president stated publicly that rather than extending the old agreement, the United States should work towards a new, improved, and modernised treaty. The following day, a senior American official told the Conference on Disarmament in Geneva that Washington would seek to bring China into arms control discussions and to limit all warheads rather than only deployed strategic warheads. Moscow's position settled into conditional restraint. In February 2026 the Russian foreign minister told the Russian parliament that Russia was observing a moratorium on exceeding the quantitative limits, that it believed the United States was in no hurry to exceed them either, and that Russian restraint would last only as long as American restraint did. So the position in mid 2026 is this. There is no treaty. There are two unilateral, conditional, unverifiable declarations of restraint, each explicitly contingent on the other. There are no negotiations. There is an American aspiration to a broader agreement including China, and a Chinese refusal to participate on the grounds that such talks would shift the burden of disarmament onto a state with a much smaller arsenal. There is a Russian statement that the conditions for a substantive strategic dialogue are, for now, absent. This is the condition that may fairly be called the #post_New_START era. 10.4 The multilateral echo The failure had an immediate multilateral consequence. The eleventh review conference of the NPT met in New York between late April and 22 May 2026 and ended without a consensus outcome document, the third consecutive review conference to do so. Non nuclear weapon states pressed the recognised nuclear weapon states on their obligations under #Article_VI, which requires good faith negotiations towards ending the arms race and towards #nuclear_disarmament. The nuclear armed states resisted specific, time bound commitments. Russia and China criticised the United States for failing to pursue a successor to New START. The United States focused its criticism on Chinese expansion and on allegations concerning nuclear testing. The dispute that formally ended the conference concerned another state's programme entirely, but the underlying picture was clear enough: the #treaty_expiry had removed the most tangible evidence that the two largest arsenals were being restrained at all. Independent research published in June 2026 reported that global arsenals had been upgraded and, in terms of active stockpiles and deployed warheads, had grown during 2025, and that every nuclear armed state was engaged in some form of #nuclear_modernization. Whatever else is true, the direction of travel is now upward. This is the environment that supporters of the #TPNW, the treaty prohibiting nuclear weapons, cite when they argue that incremental bilateral arms control has run out of road. 11. Discussion: An Assessment of the Treaty 11.1 What New START achieved Four achievements can be defended on the evidence. First, it delivered verified reductions. Both parties reached the central limits by the 2018 deadline and stayed at or below them. The deployed strategic forces of both states fell to levels not seen since the early 1960s. This was accomplished without either side losing confidence in its own deterrence. Second, it produced knowledge. For nine years the treaty generated a continuous, checkable picture of the other side's strategic forces. The value of this is invisible when it is present and obvious when it is gone. Intelligence services must now reconstruct through #open_source_monitoring, #satellite_imagery, and inference what they previously received in writing twice a year. The uncertainty band around every estimate widens, and wider uncertainty is the raw material of worst case planning. Third, it protected national technical means. The obligation not to interfere with the other side's monitoring assets, and not to use concealment measures against them, quietly supported the early warning architecture on which crisis decision making depends. As command and control systems become more digital, this protection may matter more, not less. Fourth, it preserved a working channel. The consultative commission gave technical officials from two hostile states a reason to sit in the same room and argue about launcher conversion procedures. Channels of this kind are cheap to maintain and very expensive to rebuild. 11.2 What New START failed to do Three failures are equally clear. The first failure is scope. By counting only intercontinental range systems and only deployed warheads, the treaty left the majority of both arsenals untouched. Reserve warheads, non strategic nuclear weapons, and the exotic new systems were all outside it. A treaty that constrains posture but not capability is fragile by construction, because the capability can be reconstituted whenever the political constraint weakens. That is exactly what is now feared. The second failure is adaptability. The treaty had a fixed term and a single extension, and it contained no mechanism for updating its definitions as technology changed. When Russia unveiled systems that did not fit the 2010 categories, there was no agreed process to decide whether they were accountable, only argument. A treaty that cannot be amended in the face of new technology will eventually be overtaken by it. The third failure is political insulation. New START was never separated from the wider relationship, and when that relationship collapsed, the treaty collapsed with it. The classical theory of nuclear arms control holds that adversaries can cooperate on survival even while competing everywhere else. The record of 2022 to 2026 suggests that this insulation is much harder to sustain in practice than in theory. Verification requires people to travel, to be granted visas, and to enter sensitive facilities in a country their own government is actively confronting. There may be a level of hostility beyond which intrusive verification is simply not politically survivable, and if so, that is a sobering finding for the whole field. 11.3 Was observing the old limits in the American interest? This question deserves fair treatment, because it is genuinely contested and because students will encounter strong claims on both sides. The case against observing the old limits runs as follows. Russian behaviour cannot be verified. Russia has, on this reading, failed to honour a series of arms control commitments over two decades. The security environment has changed fundamentally, because the United States must now hold at risk the forces of two large nuclear powers rather than one, and its targeting requirements have grown accordingly. The old ceilings were calibrated for a different problem, and accepting them now would constrain American forces while leaving Russian reserve warheads and the entire Chinese arsenal unconstrained. Nor would Russia's proposal cover its newest systems, including Poseidon and Burevestnik. On this view, the sensible course is to let the treaty go, prepare to expand if necessary, and negotiate only from a position of strength. The case for observing the old limits runs as follows. The United States cannot in fact expand quickly. Reconverting launchers and uploading warheads takes time and money, and the American weapons infrastructure faces well documented difficulties. Russia, with active production lines and many partially loaded MIRV missiles, may be able to move faster in the short term. Abandoning the ceilings therefore risks handing Moscow a temporary numerical advantage in exchange for a freedom the United States cannot immediately use. Meanwhile, an unconstrained American buildup gives Beijing every reason to accelerate its own, which makes the two peer problem worse rather than better. A one year observance costs little, preserves a floor of #predictability, and creates a platform for talks. Both arguments are internally coherent. They differ mainly in their estimate of two things: how fast each side can actually add deployed warheads, and whether Chinese expansion is responsive to American behaviour or independent of it. Neither of these is settled, and any student writing on this topic should say so rather than adopting one side's certainty. 11.4 The arms race question It is tempting to describe the treaty expiry as the starting gun of a new arms race. This is probably too simple, and the reason is worth understanding. The competition was already under way. All three major nuclear powers were modernising before the treaty lapsed. Chinese expansion was proceeding at a pace unrelated to the bilateral treaty, because China was never a party to it. Russian modernisation was largely complete. American modernisation was already funded, in part as the political price of the treaty's own approval in the Senate in 2010. What the expiry changes is not the existence of competition but its character. It removes the ceiling, it removes the information, and it removes the shared vocabulary. Competition without ceilings and without information is more dangerous than competition with them, because each side must now plan against the worst plausible estimate of the other rather than against a known figure. The likely trajectory is not a sprint but a long, expensive, uncertain climb, punctuated by moments in which one side misreads the other. That is a more accurate and, in some ways, a more troubling picture than a simple race. 12. Strategic Implications After Expiry 12.1 The upload problem The most immediate technical question in the post New START world is how quickly each side could raise its deployed warhead numbers if it decided to. Both arsenals were deliberately built with slack. The United States keeps a large reserve of warheads in storage, sometimes described as the hedge stockpile. Its submarine launched missiles carry fewer warheads than their maximum capacity. Some of its launch tubes and bombers were converted to non nuclear status in order to meet the New START ceilings, and at least some of those conversions could be reversed. Independent analysts have estimated that a full sea based upload might in theory add many hundreds of warheads, but that a realistic upload, one that preserves targeting flexibility rather than forcing commanders to fire everything at once, would produce a smaller figure, in the range of several hundred. Any such move would take months, because each submarine would have to return to port in turn. Russia has a comparable but differently shaped potential. Its land based force includes missiles capable of carrying multiple warheads that are not currently loaded to capacity, and it maintains a substantial reserve of stored warheads. The exact figures are estimates, and after the end of data exchanges those estimates rest on open source monitoring rather than declared data. The key insight is that breakout and reconstitution are not the same thing. Breakout means racing beyond a limit in secret. Reconstitution means openly using capacity that was always there. What both states now possess is reconstitution capacity, and the New START ceiling was never a barrier to it. The ceiling was a political decision not to use it. That decision is now unprotected by law. 12.2 The two peer problem and its logical trap The American argument that two adversaries require more weapons than one has intuitive appeal, but it contains a trap that deserves careful attention. If the United States increases its deployed force in response to China, Russia will read that increase as directed at Russia, because Russian planners cannot see intentions, only capabilities. Russia may then increase its own force. China, observing a larger American force, may accelerate further. Each state can describe its own action as defensive and each will be sincere. This is the classical security dilemma, and the absence of transparency mechanisms makes it sharper, because there is now no agreed channel through which any of the three can credibly explain what it is doing or why. The counter argument is that Chinese expansion appears to be driven by Chinese strategic goals rather than by American numbers, and that it proceeded steadily throughout the period when American deployed forces were capped. If Beijing is not reacting to Washington, then American restraint buys nothing and American expansion costs nothing in terms of Chinese behaviour. The empirical question of which story is true is, at present, unresolved, and it is the single most consequential open question in the field. 12.3 Trilateral arms control and why it is hard The stated American goal is #trilateral_arms_control, or at least a framework that captures China. Four obstacles stand in the way. The first is asymmetry. China has a far smaller arsenal than either the United States or Russia. Equal ceilings would require the two larger powers to cut deeply or would permit China to grow substantially. Neither outcome is politically easy in Washington or Moscow, and Beijing has repeatedly declined to accept limits that would freeze it in a position of inferiority. The second is Chinese refusal. Chinese officials have said plainly that they see such proposals as an attempt to transfer the responsibility for nuclear disarmament from the states with the largest arsenals to a state with a much smaller one. Until that perception changes, participation is unlikely. The third is the widening circle. Russia has insisted for years that any multilateral framework must include the British and French arsenals. Some American officials have suggested that all nuclear armed states should eventually be involved. Each addition multiplies complexity, and complexity is the enemy of completion. New START, a bilateral treaty running to hundreds of pages, took roughly a year to negotiate, and that was considered fast. The fourth is verification. A trilateral verification regime would require intrusive access in three countries with three different security cultures, and China has no experience of hosting such inspections. None of this makes multilateral arms control impossible. It does make it slow, which matters, because the bilateral floor has already been removed. 12.4 Risk reduction as a fallback Where formal treaties are unavailable, the literature increasingly turns to #risk_reduction and #confidence_building_measures. These are less ambitious than treaties and correspondingly easier to agree. The most obvious surviving example is the 1988 agreement on notifying strategic ballistic missile launches in advance, which both states have continued to observe even through the worst of the current period. A similar arrangement exists between Russia and China. It is at least conceivable that these separate bilateral arrangements could be woven into a wider multilateral notification network, which would be a modest but real gain. Other candidate measures include mutual restraint on nuclear testing, agreed norms against interference with early warning satellites, dialogue on the role of artificial intelligence in nuclear command and control, and reciprocal commitments regarding the sea based systems that are hardest to track. None of these caps a single warhead. All of them reduce the probability that an accident or a misreading becomes a catastrophe. The honest assessment is that these measures are second best. They are what remains when the first best option is gone. 13. Policy Options for a Successor Framework Based on the analysis above, five design principles can be identified for any future arrangement. These are offered as analytical propositions rather than as advocacy. First, count warheads rather than launchers. The central lesson of the hedge stockpile is that limiting deployed systems does not limit destructive capacity. Any successor that repeats the deployed only formula will inherit the same fragility. A total warhead ceiling is far harder to verify, because warheads in storage are small and easy to hide, but the verification research community has spent two decades on exactly this problem and the technical groundwork exists. Second, build in adaptability. The failure of New START to accommodate new systems such as Poseidon and Burevestnik shows the cost of fixed definitions. A successor should contain an agreed procedure for classifying new technologies, so that novelty produces a decision rather than a deadlock. Third, decouple verification from the political weather. If inspections can be blocked whenever relations sour, then verification is available only when it is least needed. Options worth study include remote and automated monitoring, third party or international inspection teams, and continuous technical monitoring that does not require inspectors to fly into hostile territory. Fourth, separate the arms control track from the wider dispute. This was the founding premise of the field and it was abandoned in practice after 2022. Whether it can be restored is uncertain, but the alternative, in which strategic restraint is held hostage to every other quarrel, has now been tested and has failed. Fifth, accept partial and informal steps. The precedent of the early 1980s, when the parties observed an unratified agreement, suggests that political commitments can hold in the absence of legal ones. A parallel unilateral declaration of ceilings, even without inspection, is worth more than nothing, particularly if paired with a resumption of launch notifications and a restored strategic dialogue. None of these will be achieved quickly. The negotiation of New START took a year in conditions of relative goodwill, with experienced teams on both sides and a clear precedent to build on. In 2026 the goodwill is absent, the expert bureaucracies in both capitals have been reduced, and the precedent is a treaty that ended in mutual recrimination. Realism requires acknowledging that the interregnum may be long. 14. Limitations of the Study Three limitations should be stated plainly. First, the evidentiary base is incomplete. The most important facts about both arsenals are classified. Everything written here about warhead numbers, upload potential, and reconstitution timelines rests on official aggregate declarations, which ceased in 2023, and on open source estimates, which are careful but uncertain. Any reader who takes a specific number from this article and treats it as exact has misread it. Second, the analysis of very recent events is provisional. The treaty expired in February 2026, and the meaning of that expiry will look different in five years. It is possible that a successor framework emerges more quickly than currently seems likely. It is equally possible that the restraint currently being observed collapses. Historical judgement is not available to the historian of the present. Third, the article draws mainly on English language material. Russian and Chinese internal debates are represented here largely through the interpretations of Western analysts, and this inevitably flattens them. A fuller study would engage directly with Russian and Chinese scholarship, and students in a position to do so should. 15. Conclusion New START was, in the end, a machine for turning fear into arithmetic. Its ceilings were significant: 1,550 deployed strategic warheads, 700 deployed delivery vehicles, 800 total launchers, achieved by both parties and maintained for the life of the treaty. Its real contribution, however, was not the ceilings. It was the eighteen inspections a year, the two annual data exchanges, the unique identifiers on the missiles, the notifications flowing quietly between capitals, and the commitment not to blind the other side's satellites. These are the things that allowed two governments that trusted each other very little to nonetheless know a great deal about each other. That knowledge is the raw material of strategic stability. It is why the treaty deserves to be studied as an institution and not merely as a set of numbers. The treaty's weaknesses were equally real. It captured a shrinking share of the nuclear problem. It could not be amended to meet new technology. It could not be insulated from the collapse of the wider relationship. It counted a bomber as one warhead when a bomber carries many. It said nothing about the thousands of warheads sitting in storage, and nothing at all about the third state whose arsenal was growing fastest. The verification regime died in 2023. The treaty itself expired on 5 February 2026. What remains is a pair of conditional promises, unverifiable and revocable, offered by two governments that no longer trust each other and no longer meet to discuss it. Restraint continues, for now, because both sides currently judge restraint to be in their interest. That is a thinner foundation than a treaty, and everyone involved knows it. The lesson for students of #international_security is not that arms control failed. It is that arms control is an infrastructure, and infrastructure requires maintenance. START I built the machinery. New START inherited and improved it. Nobody built the successor, and when the political storm came, there was nothing left to shelter in. The task now is to rebuild, and rebuilding is always harder and slower than maintaining would have been. References Arms Control Association (2026) New START at a Glance. Fact sheet. Washington, DC: Arms Control Association. Arms Control Today (2025) 'Life After New START: Navigating a New Period of Nuclear Arms Control', Arms Control Today, 55(1), January. Atlantic Council (2026) Is Extending the New START Limits in the US National Security Interest? Issue Brief. Washington, DC: Scowcroft Center for Strategy and Security, Atlantic Council. Bollfrass, A. K. and Herzog, S. (2022) 'The War in Ukraine and Global Nuclear Order', Survival, 64(4), pp. 7 to 32. Brookings Institution (2026) What Comes After New START? Commentary. Washington, DC: Brookings Institution. Bugos, S. (2021) 'U.S., Russia Extend New START for Five Years', Arms Control Today, 51(2), March. Bulletin of the Atomic Scientists (2026) Doomsday Clock Statement. Chicago, IL: Bulletin of the Atomic Scientists Science and Security Board. Congressional Research Service (2021) The New START Treaty: Central Limits and Key Provisions. CRS Report R41219. Washington, DC: Congressional Research Service. Congressional Research Service (2022) Russia's Nuclear Weapons: Doctrine, Forces, and Modernization. CRS Report R45861. Washington, DC: Congressional Research Service. Congressional Research Service (2026) U.S.-Russian Nuclear Arms Control: Overview and Potential Considerations for Congress. CRS In Focus IF12964. Washington, DC: Congressional Research Service. Council on Foreign Relations (2026) Nukes Without Limits? A New Era After the End of New START. New York: Council on Foreign Relations. Federation of American Scientists (2026) The Aftermath: The Expiration of New START and What It Means For Us All. Washington, DC: Federation of American Scientists. Finnish Institute of International Affairs (2026) Expiration of New START. FIIA Comment. Helsinki: Finnish Institute of International Affairs. Flatoff, L. and Kimball, D. G. (2026) '2026 NPT Review Conference Stymied by Disputes', Arms Control Today, 56(5), June. Futter, A. (2021) The Politics of Nuclear Weapons. New edition. Cham: Palgrave Macmillan. Gottemoeller, R. (2021) Negotiating the New START Treaty. Amherst, NY: Cambria Press. International Campaign to Abolish Nuclear Weapons (2026) The Expiration of New START: What It Means and What Is Next. Geneva: ICAN. Klare, M. T. (2026) 'What Will the End of the US and Russia Nuclear Arms Treaty Mean?', Inkstick Media, March. Kristensen, H. M., Korda, M., Johns, E. and Knight, M. (2024) 'Russian Nuclear Weapons, 2024', Bulletin of the Atomic Scientists, 80(2), pp. 118 to 145. Kristensen, H. M., Korda, M., Johns, E. and Knight, M. (2024) 'United States Nuclear Weapons, 2024', Bulletin of the Atomic Scientists, 80(3), pp. 182 to 208. Liang, X. (2026) 'New START Expires as U.S. Urges Modernized Treaty', Arms Control Today, 56(2), March. Narang, V. and Sagan, S. D. (eds.) (2022) The Fragile Balance of Terror: Deterrence in the New Nuclear Age. Ithaca, NY: Cornell University Press. Royal United Services Institute (2025) Beyond New START: What Happens Next in Nuclear Arms Control. RUSI Commentary. London: Royal United Services Institute. Sokov, N. (2026) End of New START: Short and Medium Term Options. Vienna: Vienna Center for Disarmament and Non-Proliferation. Stockholm International Peace Research Institute (2024) SIPRI Yearbook 2024: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Stockholm International Peace Research Institute (2025) SIPRI Yearbook 2025: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Stockholm International Peace Research Institute (2026) SIPRI Yearbook 2026: Armaments, Disarmament and International Security. Oxford: Oxford University Press. United Nations Office for Disarmament Affairs (2026) Eleventh Review Conference of the Parties to the Treaty on the Non-Proliferation of Nuclear Weapons: Final Documents and Statements. New York: United Nations. United States Department of Defense (2024) Military and Security Developments Involving the People's Republic of China: Annual Report to Congress. Washington, DC: Office of the Secretary of Defense. United States Department of State (2025) New START Treaty. Fact sheet. Washington, DC: Bureau of Arms Control, Deterrence and Stability. Hashtags #NewSTART #NewSTARTTreaty #New_START_2010 #NuclearArmsControl #StrategicArmsReduction #USRussiaNuclearRelations #DeployedNuclearWarheads #StrategicStability #NuclearVerification #NuclearNonproliferation #ArmsControlAfterNewSTART #NuclearDeterrenceStudies #TreatyExpiration2026 #DisarmamentStudies #InternationalSecurityResearch

The Strategic Arms Reduction Treaty, signed in Moscow on 31 July 1991 by the United States and the Soviet Union, was the first agreement in the nuclear age that required both superpowers to destroy strategic offensive weapons rather than simply agree not to build more of them. This article re-examines #START_I as a legal instrument, a verification system, and a political achievement, and asks what it can still teach students of international security in a period when the last bilateral cap on American and Russian strategic forces has lapsed. The study uses a qualitative, document-centred method, combining treaty analysis, historical process tracing, and a review of scholarship published between 2021 and 2026. Four findings emerge. First, the treaty's real innovation was not the numbers it set but the machinery it built: counting rules, data exchanges, telemetry access, and twelve categories of #on_site_inspection that turned promises into observable behaviour. Second, the treaty survived the collapse of one of its own parties, and its adaptation through the 1992 #Lisbon_Protocol converted a bilateral bargain into a four-state settlement that helped denuclearise Belarus, Kazakhstan, and Ukraine. Third, the reductions were substantial but narrower than public memory suggests, because the treaty limited accountable deployed warheads and delivery vehicles rather than total stockpiles, non-strategic weapons, or warheads held in reserve. Fourth, the political conditions that made #START_I possible, including leadership continuity, technical trust built through the earlier intermediate-range treaty, and a shared interest in predictability, are precisely the conditions that are missing today. The article concludes that #START_I remains the most complete working model of #verification-based #nuclear_arms_control ever implemented, and that any future framework, bilateral or multilateral, will most likely be assembled from its parts rather than invented from nothing. Keywords: Strategic Arms Reduction Treaty; nuclear arms control; verification; strategic stability; Lisbon Protocol; denuclearisation; New START; deterrence 1. Introduction For most of the Cold War, the United States and the Soviet Union negotiated about nuclear weapons in order to manage the growth of their arsenals rather than to shrink them. The Strategic Arms Limitation Talks of the 1970s produced ceilings, freezes, and definitions. They did not produce a single destroyed missile that would not otherwise have been retired. The Strategic Arms Reduction Treaty of 1991 broke that pattern. It obliged both parties to cut their #strategic_offensive_arms to agreed levels within seven years, to prove that they had done so, and to let the other side watch. That difference between limitation and reduction is not a technicality. It changed what an arms control agreement was for. A limitation treaty asks a state to promise restraint in the future. A reduction treaty asks a state to give up capability it already possesses, in a way that the adversary can see and count. The second demand is far harder politically, because it touches military services, industrial constituencies, and national prestige. It is also far harder technically, because it requires agreement on what counts as a weapon, where it may be located, how it will be destroyed, and how destruction will be confirmed. #START_I answered all of these questions in a document that ran to hundreds of pages of treaty text, protocols, annexes, and a memorandum of understanding containing detailed data on every declared facility. The treaty is worth studying today for three reasons that go beyond historical interest. The first is that it worked. The parties reached the central limits, and the inspection regime functioned for fifteen years across an extraordinary political transition, including the disappearance of one signatory state. Very few security agreements can claim that record. The second is that it is the direct ancestor of everything that followed. START II, the 2002 Moscow Treaty, and the 2010 New START agreement all borrowed from its definitions, its counting rules, and its inspection procedures. When the 2002 treaty was signed with no verification annex of its own, it could survive only because the #verification system of #START_I was still running underneath it. This is a striking illustration of how institutional machinery, once built, becomes a public good that later agreements consume. The third reason is the most urgent. On 5 February 2026 the New START treaty expired, and for the first time since the early 1970s there is no legally binding ceiling on the strategic nuclear forces of the United States and the Russian Federation. Both governments have signalled interest in some form of restraint, and Russia offered in September 2025 to observe the old numerical limits for a further year without inspections, but no successor instrument is in force. In this environment, the question of what a working #nuclear_arms_control regime actually requires has moved from the seminar room back to the policy agenda. This article therefore has four aims. It reconstructs the negotiating history of the treaty, explains its legal and technical architecture in accessible language, evaluates its implementation and outcomes against the claims made for it, and draws out lessons for the current period. It is written for advanced undergraduate and postgraduate students who need a single reliable account of a treaty that is often mentioned in textbooks and rarely explained. The argument advanced here is that #START_I should be understood less as a disarmament agreement and more as an information system with teeth. Its lasting contribution was to demonstrate that two hostile states with tens of thousands of nuclear weapons between them could construct a shared factual picture of each other's forces, and could act on that picture without either side losing confidence in its own security. Whether that demonstration can be repeated is now an open question. 2. Literature Review Scholarship on the treaty falls into four broad streams, and each has been revisited in the past five years. The first stream is the diplomatic history of the late Cold War. Recent work has moved away from a simple story in which American pressure produced Soviet concessions, and towards accounts that give weight to internal Soviet reassessment, to alliance politics inside NATO, and to the personal role of negotiators. Colbourn's study of the intermediate-range missile crisis in Europe shows how the deployment controversy of the early 1980s nearly broke the Atlantic alliance and how its resolution created both the political appetite and the technical vocabulary for deeper cuts (Colbourn 2022). Plokhy's re-examination of the Cuban crisis, though set two decades earlier, matters here because it documents how thin the informational basis for crisis decision-making was before any inspection regime existed (Plokhy 2021). Read together, these histories support a view of #START_I as the product of a long learning process rather than a sudden breakthrough. The second stream is the study of #verification and #irreversibility. This literature has grown quickly since 2023, partly because the collapse of existing agreements has forced analysts to ask what would have to be rebuilt. Woolf's account of the elimination procedures used in the US-Soviet and US-Russian process is the most directly relevant recent contribution, because it explains how the physical destruction of launchers and missiles was specified in treaty text and why that specification made reversal costly (Woolf 2024). Fleck approaches the same problem from international law and argues that the legal norms supporting #irreversibility are real but full of gaps, since the treaties are not universal, missile limits have been abandoned, and doctrines remain opaque (Fleck 2023). Rodgers and Williams identify a paradox at the centre of the field: negotiators want agreements that are hard to undo, but states want flexibility in case the security environment worsens, so every agreement contains a built-in escape route (Rodgers and Williams 2023). Elbahtimy frames #irreversibility as a political problem before it is a technical one, and calls for empirical study of cases where reductions actually stuck (Elbahtimy 2023; Elbahtimy 2024). Ritchie extends the argument from weapons to the industrial complexes that produce them, noting that a warhead can be rebuilt if the factory survives (Ritchie 2023). Pretorius adds the South African case, the only instance of a state dismantling an indigenous arsenal, and stresses that the durability of that decision rested on domestic political transformation rather than on external monitoring alone (Pretorius 2024). The third stream concerns the post-Soviet nuclear settlement. Budjeryn's archival history of Ukrainian #disarmament is now the standard work, and it substantially revises the older picture in which Kyiv simply handed over weapons it could not use. She shows that Ukrainian officials understood the strategic value of the inherited arsenal, that they negotiated hard over compensation and security assurances, and that the eventual transfer to Russia reflected a calculated judgement about sovereignty, economic survival, and international recognition rather than passive compliance (Budjeryn 2023). Since 2022 this literature has acquired an uncomfortable second life, because the fate of the #Budapest_Memorandum has become a standard argument in proliferation debates worldwide. The fourth stream is contemporary strategic analysis. The annual Nuclear Notebook estimates provide the most reliable open data on current forces and confirm that both the United States and Russia are engaged in comprehensive #nuclear_modernization programmes that will replace every leg of their triads (Kristensen, Korda, Johns, and Knight 2024; Kristensen, Korda, Johns, and Knight 2025). Kristensen, Korda, and Rogers set out the legal and political options available after the New START treaty and warn that the window for negotiated restraint was closing well before the expiry date (Kristensen, Korda, and Rogers 2022). Narang and Sagan's edited volume argues that the assumptions of the bipolar era, including the belief that survivable second-strike forces are easy to maintain, no longer hold in a world of precise conventional weapons, missile defences, and multiple nuclear rivals (Narang and Sagan 2022). Cunningham's study of Chinese capabilities and Johnson's analysis of artificial intelligence in nuclear command and control both point in the same direction: the technical environment that #START_I was designed for has changed (Cunningham 2025; Johnson 2023). A gap remains. The verification literature treats #START_I mainly as a source of technical precedent, and the historical literature treats it mainly as a diplomatic endpoint. There is relatively little work that reads the treaty as a whole system, in which the political bargain, the counting rules, and the inspection procedures were mutually dependent. This article attempts that integrated reading. 3. Conceptual Framework Three concepts organise the analysis. Strategic stability. In the classical formulation, a nuclear relationship is stable when neither side has an incentive to strike first in a crisis and neither has an incentive to build up its forces rapidly in peacetime. The first condition is crisis stability; the second is arms race stability. Weapons that are accurate, numerous, and vulnerable undermine both, because they are attractive targets and useful attackers. Silo-based missiles carrying many independently targetable warheads are the classic example, since a single incoming warhead can destroy several enemy warheads on the ground. The whole design of #START_I can be read as an attempt to reduce this problem, which is why it capped ballistic missile warheads more tightly than bomber weapons and why it singled out the largest Soviet heavy missiles for special treatment. Recent scholarship questions whether #strategic_stability retains its analytical value in a multi-actor system, but it remains the standard against which the treaty was designed and should be judged (Narang and Sagan 2022; Mount 2025). Verification and the management of distrust. #Verification does not create trust; it makes trust unnecessary for limited purposes. A well-designed regime allows each party to detect militarily significant cheating in time to respond, and allows each party to demonstrate its own compliance credibly. Both functions matter. The second is often overlooked, yet it is what makes an agreement politically survivable at home, because it protects the compliant state from accusations that it has been deceived. #START_I combined national technical means, meaning satellites and other remote sensors, with intrusive on-site methods that could see what satellites cannot, such as the number of warheads on a missile inside a silo (Woolf 2024; Roehrlich 2022). Irreversibility. A reduction is irreversible to the extent that reversing it would be slow, expensive, visible, or politically costly. Perfect irreversibility is unattainable, because knowledge cannot be destroyed and industrial capacity can be rebuilt. What treaties can do is raise the cost and the visibility of reconstitution. The distinction between eliminating a warhead and eliminating the launcher that carries it is central here, and it is a distinction that #START_I did not fully bridge: the treaty counted warheads but destroyed delivery systems (Woolf 2024; Ritchie 2023; Rodgers and Williams 2023). These three concepts generate the questions that structure the empirical sections. Did the treaty improve #strategic_stability, or merely reduce numbers? Did its #verification system actually detect and deter non-compliance, or was it symbolic? And did its reductions prove durable once the political conditions that produced them disappeared? 4. Research Design and Method This is a qualitative, single-case study using document analysis and structured historical reconstruction. Three categories of material are used. The first is the treaty corpus itself, meaning the treaty text, its protocols on conversion and elimination, inspections, notifications, and telemetry, the memorandum of understanding on data, and the 1992 #Lisbon_Protocol. These documents are read for their legal structure and for the technical definitions they establish. The second is the secondary scholarly literature published between 2021 and 2026, selected for peer-reviewed status and topical relevance. Preference is given to work in specialised journals of arms control, security studies, and international law, and to monographs from established academic presses. The third is authoritative technical estimation, principally the open-source assessments of national nuclear forces published annually by specialist research institutes, used here for force-level data rather than for interpretation. The method is process tracing. Rather than testing a general hypothesis across many cases, the analysis follows the causal chain from negotiating positions to treaty text to implementation behaviour to outcomes, and asks at each step whether the observed result is best explained by the treaty or by other factors, such as budgetary collapse in the former Soviet Union. Two limitations should be stated at the outset. Archival access on the Soviet and Russian side remains uneven, so accounts of internal decision-making rely heavily on memoir and on the small body of documents that has been released. And because there is only one case, the findings cannot be generalised statistically; they are offered as analytically generalisable propositions about what verified reduction requires. 5. The Road to 1991: From Limitation to Reduction 5.1 The inheritance of the 1970s By the end of the 1970s the two superpowers had built a strategic relationship that neither could control. Each had thousands of deployed warheads, and the introduction of multiple independently targetable re-entry vehicles meant that a single missile could carry ten or more of them. The arithmetic of the #arms_race turned vicious: because one attacking warhead could in principle destroy several defending warheads sitting in a silo, the side that struck first appeared to gain. Nobody believed a first strike would actually be attempted, but the structure of the problem drove procurement on both sides. The agreements of that era did not solve this. They set ceilings high enough to accommodate planned forces, and they lacked the intrusive #verification needed to look inside a missile. The result was a series of compliance disputes that poisoned the domestic politics of arms control in Washington and gave critics an easy argument: that the Soviet Union cheated and that treaties tied American hands. 5.2 Reagan's proposal and the collapse of 1983 When the Reagan administration reopened negotiations in 1982, it renamed the process to signal a change in ambition. The talks would no longer be about limitation but about reduction. The initial American position asked for deep cuts concentrated in exactly the area where the Soviet Union was strongest, namely large land-based missiles with many warheads. Moscow saw this as an attempt to disarm it unilaterally while leaving American bombers and submarines untouched. The talks collapsed in late 1983, when the Soviet delegation walked out in protest at the deployment of American intermediate-range missiles in Western Europe. That crisis, and the fear it generated on both sides, is now understood as a turning point rather than a simple breakdown (Colbourn 2022). The near-miss atmosphere of 1983 and 1984 convinced leaders in both capitals that the risk of inadvertent war had become intolerable. 5.3 The intermediate-range treaty as a technical laboratory Talks resumed in 1985. The change of leadership in Moscow mattered, but so did a change of method. The 1987 treaty on intermediate-range forces eliminated an entire class of missiles and, more importantly for our purposes, introduced #on_site_inspection and continuous portal monitoring at a missile production plant. For the first time, American inspectors stood at the gate of a Soviet factory and counted what came out, and Soviet inspectors did the same in the United States. This is the single most underrated precondition for #START_I. The intermediate-range treaty proved that intrusive inspection was survivable. It created procedures, trained personnel, and generated a body of shared experience. When negotiators later argued about how to verify warhead numbers on strategic missiles, they were arguing within a framework that already existed. 5.4 The endgame, 1989 to 1991 The final phase was shaped by three developments. The Soviet economy was deteriorating, which made the maintenance of an enormous strategic force increasingly unaffordable. The political ground in Eastern Europe shifted decisively in 1989. And both militaries began, quietly, to see benefit in a treaty that would cap the other side's forces during a period of internal turmoil. Difficult issues remained until the end: how to count weapons carried by heavy bombers, how to handle mobile land-based missiles that could hide in forests, how to treat sea-launched cruise missiles, and whether telemetry from test flights could be encrypted. The eventual solutions were compromises, and several of them were frankly artificial. But the treaty was signed in Moscow on 31 July 1991, less than three weeks before an attempted coup in the Soviet capital and five months before the state that signed it ceased to exist. 5.5 The two problems that almost stopped the treaty Two disputes deserve attention because they show how technical questions become political ones. The first was mobile land-based missiles. The Soviet Union had developed missiles carried on wheeled launchers and on railway carriages, which could disperse across an enormous territory and hide in forests or tunnels. From an American perspective these weapons were a verification nightmare, because satellites cannot reliably count objects that move and hide, and a mobile force could in principle be far larger than declared. Some American negotiators wanted them banned outright. From a Soviet perspective they were precisely the systems that guaranteed a surviving retaliatory force, and therefore contributed to crisis stability. Banning them would have pushed Moscow towards fixed silos that invite a first strike. The compromise reveals the treaty's method. Mobile missiles were permitted but limited, both in warhead numbers and by rules that restricted where they could be based and how far they could roam. They were subjected to specific transparency requirements, including limits on the number of launchers that could be outside their declared bases at any moment, restrictions on concealment, and continuous monitoring of the plant where they were assembled. The problem was not solved; it was made observable, which for arms control purposes is often the same thing. The second dispute was sea-launched cruise missiles. These are small, can be carried on many types of ship or submarine, and are almost impossible to count from outside a vessel's hull. Including them would have required a level of naval intrusion neither navy would tolerate. Excluding them would have left an obvious loophole. The eventual answer was to take them out of the treaty altogether and handle them through parallel political declarations in which each side stated the number of such weapons it intended to deploy. These declarations were politically binding but legally separate and not verifiable. That outcome is instructive in a negative way. It shows the boundary of what the parties could achieve, and it left a category of weapons outside the regime that has never subsequently been captured by any agreement. 6. The Architecture of the Treaty 6.1 The central limits The treaty imposed a set of nested ceilings, to be reached in three phases over seven years from entry into force. Each party was permitted no more than 1,600 deployed strategic nuclear delivery vehicles, a category that included #ICBMs, #SLBMs, and #heavy_bombers. Attributed to those delivery vehicles, each party could hold no more than 6,000 accountable warheads. Within that overall figure sat the provisions that carried most of the strategic logic. No more than 4,900 of the accountable warheads could be carried on ballistic missiles, which is where the first-strike problem lived. No more than 1,100 could be carried on mobile land-based missiles. And the Soviet side was limited to 154 heavy #ICBMs carrying no more than 1,540 warheads, which required the destruction of half of the existing force of very large missiles that American planners regarded as the most destabilising weapons in the world. A separate constraint capped the aggregate throw-weight of ballistic missiles, a measure of the total payload that could be lifted, at 3,600 metric tonnes. 6.2 Counting rules and their fictions The most instructive feature of the treaty for students is that it did not count actual warheads. It counted attributed warheads, using rules agreed in advance. Each type of ballistic missile was assigned a number of warheads. That number was fixed for treaty purposes regardless of how many warheads a particular missile carried on a particular day. This was necessary because opening a missile's nose cone to count warheads is not something either side was prepared to allow routinely, but it created a gap between treaty accounting and physical reality. Bombers were treated even more loosely. A heavy bomber not equipped for long-range nuclear cruise missiles counted as a single warhead, even though it could carry many gravity bombs. Bombers equipped for cruise missiles were counted using a discount: a fixed attribution well below actual carrying capacity applied up to a specified number of aircraft on each side, with the real figure applying thereafter. The effect was to allow both air forces to keep more weapons than the #warhead_limits appeared to permit. Two justifications were offered. The technical one was that bombers are slow, recallable, and vulnerable to air defences, so they are stabilising weapons that need not be counted strictly. The political one was that neither air force would accept a treaty that gutted it. Both justifications were partly true, and the result was a treaty whose headline number of 6,000 corresponded to a real deployed force closer to 8,000 or 9,000 weapons on each side. The treaty also regulated downloading, meaning the removal of warheads from a missile to reduce its accountable number. Downloading was permitted, but only within specified caps and with the requirement that the missile's platform be physically modified in ways that made rapid re-loading difficult to conceal. The purpose was to limit upload potential, that is, the ability to break out of the treaty quickly by putting the warheads back. 6.3 Elimination procedures The reduction obligations were satisfied by eliminating delivery vehicles and launchers, not warheads. Silos were destroyed by explosive demolition or excavation and left open to satellite observation for a set period. Submarine launch tubes were removed or filled. Bombers were cut into pieces at defined points and the pieces left in the open. Missiles were destroyed by burning the motors or by crushing the airframe, again under observation. This is a crucial point that students frequently miss. A warhead removed from an eliminated missile did not have to be dismantled. It could be, and in many cases was, placed in storage. The treaty reduced deployed capability and it reduced the means of delivery, but it did not by itself reduce stockpiles. The gap between delivery-system elimination and warhead dismantlement is the central unfinished business of the entire arms control tradition, and it is why current scholarship on #irreversibility keeps returning to it (Woolf 2024; Ritchie 2023). 6.4 Structure of the instrument The agreement was organised in tiers. The treaty proper set out obligations. Protocols dealt with conversion and elimination, inspection procedures, notifications, and telemetry. Annexes carried the technical detail, including facility lists and diagrams. A memorandum of understanding contained the agreed data: numbers, types, and locations of every accountable item, deployed and non-deployed, together with the coordinates and layouts of the facilities where they were held. This structure had a practical purpose. Technical detail could be updated by the implementing commission without reopening the treaty for ratification. It is a design lesson that later agreements copied. 6.5 Ratification and entry into force Signature is not consent to be bound. The American Senate gave its advice and consent in 1992, and the vote was not close, in part because the treaty was seen as locking in an advantage at a moment of Soviet weakness and in part because the inspection provisions answered the criticism that had dogged earlier agreements. The complication lay elsewhere. The treaty could not enter into force until all five parties, meaning the United States and the four post-Soviet successor states, had completed ratification, and until Belarus, Kazakhstan, and Ukraine had acceded to the non-proliferation treaty as non-nuclear-weapon states. That linkage was deliberate. It gave the international community leverage, but it also gave Kyiv leverage, because Ukraine's signature was the last one needed. Entry into force therefore came more than three years after signature, on 5 December 1994, when the instruments of ratification were exchanged in Budapest on the same day that the security assurance memoranda were signed. The sequence matters: the reduction obligations of a treaty negotiated between two superpowers were activated by a bargain struck with a state that had not existed when the treaty was written. One further point should be made about timing. Both sides began destroying systems before the treaty entered into force, partly because unilateral initiatives announced in 1991 and 1992 had already withdrawn and consolidated large numbers of weapons, and partly because delaying elimination would have made the seven-year schedule harder to meet. Compliance, in other words, started before the legal obligation did, which is a strong signal that both militaries had internalised the logic of the agreement. 7. The Verification Regime 7.1 What was agreed #Verification rested on four pillars. The first was #national_technical_means, principally reconnaissance satellites. The treaty protected them by prohibiting interference and by banning concealment measures that would impede monitoring, including a specific ban on covering missiles or silos in ways that hid their features. The second was data. The parties exchanged the full memorandum data set at fixed intervals, and were required to notify each other whenever anything changed: when a missile moved, when a bomber was converted, when a facility opened or closed. Over the life of comparable agreements the number of such notifications ran into the tens of thousands. The cumulative effect was a continuously updated shared database of the strategic forces of both states, something that had not existed at any previous point in the nuclear age. The third was #on_site_inspection. The treaty established twelve distinct categories, each designed to answer a specific question. Baseline inspections confirmed that the initial data were accurate. Data update inspections checked for drift. Reentry vehicle inspections allowed an inspector to look at the front end of a deployed missile and confirm that it carried no more warheads than the number attributed to it. New facility, suspect site, conversion or elimination, and post-exercise dispersal inspections covered the other pathways by which a party might evade its obligations. In addition, each side maintained a permanent monitoring presence at a mobile missile final assembly plant on the other's territory, watching and measuring what came out of the gate. The fourth was #telemetry. During test flights, missiles broadcast performance data. Encrypting that data would allow a party to hide the true capabilities of a missile, including how many warheads it could carry. The treaty banned encryption and jamming of telemetry associated with accountable systems and required the exchange of recordings and of the technical information needed to interpret them. This was, in some ways, the most intrusive provision of all, because it exposed engineering details that militaries normally guard closely. 7.2 How it performed In the first seven years of implementation the two sides conducted several hundred inspections between them, with the United States carrying out somewhat more than Russia. Disputes arose, as they always do, and were handled through the joint commission established for the purpose. Some concerned the interpretation of definitions, others the treatment of particular facilities. None escalated into an accusation of militarily significant cheating. Two conclusions follow. The narrow one is that the system worked as designed: it produced information, and the information was broadly consistent with declarations. The broader one is more interesting. The regime shifted the character of the relationship. Intelligence agencies on both sides gained a baseline against which to interpret ambiguous observations, which reduced the tendency towards worst-case estimation. Practitioners of the New START era later described exactly this benefit, and warned that its loss would push analysts back towards wide uncertainty ranges and therefore towards larger force requirements (Kristensen, Korda, and Rogers 2022). 7.3 The limits of the regime The system had real blind spots, and honesty about them is part of the scholarly value of the case. It did not count warheads in storage. It did not touch #non_strategic weapons, the shorter-range systems that today form the largest area of uncertainty in the Russian arsenal. It could not verify dismantlement, only elimination of launchers. It was expensive, requiring permanent inspection teams, dedicated aircraft, laboratories, and a substantial bureaucracy on each side. And it depended on a level of physical access that has since become politically unavailable. 8. The Collapse of a Party: Lisbon, Budapest, and Cooperative Threat Reduction 8.1 An unprecedented legal problem Four and a half months after signature, the state that had signed the treaty dissolved. Strategic weapons deployed on Soviet territory were now located in four independent republics: the Russian Federation, Ukraine, Kazakhstan, and Belarus. Ukraine alone held an arsenal that, counted by warhead numbers, was larger than those of the United Kingdom, France, and China combined. The legal question was novel. Was the treaty void because one party had ceased to exist? Was Russia the sole successor, entitled to inherit the obligations along with the weapons? Or were all four successor states parties to it? 8.2 The Lisbon Protocol The answer was negotiated at Lisbon on 23 May 1992. Under the #Lisbon_Protocol, Belarus, Kazakhstan, Russia, and Ukraine all became parties to the treaty as successors to the Soviet Union. The three non-Russian states undertook to accede to the nuclear non-proliferation treaty as non-nuclear-weapon states in the shortest possible time, and to remove all strategic nuclear weapons from their territory. The protocol is a small document with large consequences. It preserved the treaty by expanding it. It also linked strategic arms reduction to the global non-proliferation regime, so that the failure of one would have damaged the other. 8.3 Ukraine's decision Ratification did not follow automatically. The Ukrainian parliament attached conditions in 1993, seeking compensation for the value of the highly enriched uranium in the warheads, assistance with dismantlement costs, and, above all, security assurances. Budjeryn's archival work makes clear that Ukrainian decision-makers were not naive about what they were giving up. They understood that the weapons on their soil were not under their operational control, that maintaining them would be extremely costly, and that attempting to keep them would isolate the new state internationally. But they also understood that they were surrendering a form of insurance, and they wanted something in return (Budjeryn 2023). The trilateral statement of January 1994 unlocked the process, providing for the transfer of warheads to Russia for dismantlement, compensation in the form of nuclear fuel, and American assistance. On 5 December 1994, at Budapest, the United States, the United Kingdom, and the Russian Federation signed memoranda providing security assurances to Ukraine, Belarus, and Kazakhstan in connection with their accession to the non-proliferation treaty. The signatories committed to respect the independence, sovereignty, and existing borders of the states concerned, and to refrain from the threat or use of force against their territorial integrity or political independence. On the same day, the instruments of ratification were exchanged and the treaty entered into force. The subsequent history of the #Budapest_Memorandum is now a permanent feature of proliferation debates. The annexation of Crimea in 2014 and the full-scale invasion of 2022 constituted a breach of the commitments given in 1994 by one of the guarantor states. The analytical lesson is precise and should be stated carefully: the assurances were political commitments, not a treaty of alliance, and they contained no enforcement mechanism. States considering #disarmament now read that record, and the effect on the credibility of security assurances as a non-proliferation tool has been severe (Budjeryn 2023; Fleck 2023). 8.4 Cooperative Threat Reduction None of this could have happened without money and engineering. The legislation adopted in the United States at the end of 1991, commonly known by the names of its two Senate sponsors, created the #Cooperative_Threat_Reduction programme. It funded the transport, storage, and destruction of weapons and infrastructure across the former Soviet Union: secure railcars, storage containers, silo demolition, submarine dismantlement, the redirection of weapons scientists to civilian research. The programme is the practical answer to a question that arms control theory often ignores. A state may be willing to disarm and still be unable to do so, because #disarmament is expensive and dangerous work. The link between the treaty's legal obligations and the assistance programme's engineering capacity is one of the strongest arguments in the case that #START_I succeeded. It is also a reminder that the last strategic warheads left Ukraine in 1996, Kazakhstan in 1995, and Belarus in 1996, which was a remarkably fast timetable for an operation of this scale. 9. Implementation and Outcomes 9.1 What the numbers show At signature, each side declared strategic forces in the region of ten thousand accountable warheads, carried on well over two thousand delivery vehicles. By the deadline of 5 December 2001, both had reached the central limits of 6,000 accountable warheads and 1,600 delivery vehicles, and in practice had gone below them. The three non-Russian successor states had become non-nuclear-weapon states. Thousands of silos, submarine launch tubes, missiles, and bombers had been physically destroyed. The often-quoted claim that the treaty removed roughly eighty per cent of the world's strategic nuclear weapons deserves scrutiny. It is defensible if the baseline is the deployed strategic forces of the two superpowers at their Cold War peak and the endpoint is deployed strategic forces after full implementation and after the further reductions of the following decade. It is misleading if it is taken to mean that eighty per cent of nuclear warheads were dismantled during the treaty's implementation period. Many warheads were removed from delivery vehicles and placed in reserve rather than destroyed. The honest formulation is that the treaty, together with the unilateral initiatives of 1991 and 1992 and the assistance programmes that followed, produced the largest verified reduction in deployed #strategic_offensive_arms in history, while leaving large stockpiles of non-deployed and non-strategic weapons untouched. 9.2 The problem of attribution How much of the reduction can be credited to the treaty rather than to circumstances? This is the hardest analytical question in the case, and it should not be answered glibly. The Soviet Union and then Russia were in severe economic distress. Much of the force reduction would probably have happened anyway, because the money to maintain it was not there. The United States, for its part, was cutting defence budgets and had already begun retiring older systems. Three arguments nonetheless support a causal role for the treaty. First, the treaty determined the shape of the reductions, not only their size. Without it, each side would have retired whatever was most expensive to keep. With it, both retired what the other side found most threatening, particularly heavy #ICBMs with many warheads. That is a #strategic_stability gain that budget pressure alone would not have produced. Second, the treaty made the reductions verifiable, which meant each side could cut without fearing that the other was quietly retaining a decisive edge. Third, the treaty made the reductions difficult to reverse in the short term, because silos filled with concrete and bombers cut into sections cannot be restored quickly (Woolf 2024). The reasonable conclusion is that the collapse of Soviet power made deep cuts possible and that the treaty made them safe, structured, and durable. 9.3 The cost of compliance Arms control is often presented as a way of saving money, and in the long run it is. In the short run it is expensive, and the case makes this unusually clear. Destroying a missile silo, dismantling a ballistic missile submarine, cutting up a bomber, safely transporting a warhead across a continent, and storing fissile material under guard all cost real money and require specialised facilities. So does the inspection regime itself: escorts, transport aircraft, monitoring equipment, translators, laboratories, and a permanent staff to manage notifications and prepare data submissions. For the United States these costs were absorbed within a defence budget that was shrinking but still enormous, and a substantial portion of the assistance programme's budget went to paying for elimination work on the territory of the successor states. For those states, which were in economic freefall, the costs would have been prohibitive without help. Kazakhstan and Belarus faced the additional problem of securing sites they had never controlled and did not have the technical cadre to manage. The general lesson is that a reduction treaty is only as good as the implementation capacity behind it. A signature that a state cannot afford to honour produces non-compliance without any intention to cheat. Future frameworks that involve states with limited resources will need an equivalent of the assistance programme, and the collapse of that programme's bilateral basis after 2012 removes a tool that was central to the success of the 1990s. 10.4 A counterfactual worth considering It is worth pausing on what might have happened without the treaty and its associated protocol, because the counterfactual clarifies the stakes. Without the #Lisbon_Protocol, the legal status of the weapons in Ukraine, Kazakhstan, and Belarus would have been contested. Each new state would have faced a domestic debate about whether to keep them. In Ukraine that debate was already live, and there were serious voices arguing that an independent deterrent was the only reliable protection against a larger neighbour. The technical obstacles were real, since operational control remained with Moscow, but they were not insurmountable over a period of years. A world with four post-Soviet nuclear-armed states would have transformed the non-proliferation regime, probably fatally. Other states would have drawn conclusions. The safe passage of tens of thousands of weapons and huge quantities of fissile material through a period of state collapse, without a single confirmed diversion, is an achievement that looks more impressive with each passing year, and it was achieved through a combination of treaty obligation, financial assistance, and sustained diplomacy rather than through any one of them alone. 10. Analysis: Why It Worked, and What It Could Not Do 10.1 Conditions of success Five conditions stand out. Prior technical experience. The intermediate-range treaty had already normalised #on_site_inspection. Negotiators did not have to establish the principle of intrusive #verification and design its procedures at the same time. Alignment of military interests. Both general staffs concluded that a capped and observable adversary was preferable to an unobserved one. Arms control that runs against the professional judgement of the armed forces does not survive ratification. A long timetable. Seven years of phased reductions gave both sides time to plan force structures, retrain personnel, and absorb industrial consequences. Institutionalised dispute settlement. The joint commission converted arguments into agenda items. Without such a forum, technical disagreements become political accusations. Political leadership with domestic authority. Both leaders could deliver ratification. This is not a trivial condition, and it is one that has failed repeatedly since. 10.2 Structural weaknesses The treaty's limitations are as instructive as its achievements. It counted the wrong objects for the purpose of #disarmament, in the sense that it counted delivery vehicles and attributed warheads rather than actual weapons. It ignored non-strategic weapons entirely, leaving thousands of shorter-range warheads outside any limit, an omission that has become one of the most contentious issues in the current period. It relied on counting fictions, particularly for #heavy_bombers, that inflated real forces above the paper ceilings. It left substantial upload potential in place, because warheads removed from missiles were stored rather than destroyed. And it was profoundly bilateral, designed for a world in which two states held almost all the relevant weapons. That last point is the most consequential today. The treaty's logic assumed that limiting two arsenals against each other stabilised the system. In a world where a third arsenal is growing rapidly and where regional nuclear powers make their own calculations, this assumption no longer holds automatically (Cunningham 2025; Narang and Sagan 2022). 10.3 The information function If there is a single insight to carry away from the case, it is that the treaty's most valuable output was not reduction but information. Consider what each government knew about the other's strategic forces before 1991 and after 1994. Before, it had satellite imagery, signals intelligence, and inference. After, it had a declared database, updated continuously, checked by inspectors on the ground, and cross-referenced with telemetry from test flights. The reduction in uncertainty was enormous, and uncertainty is what drives worst-case planning. When planners must assume the maximum plausible enemy force, they request the forces needed to counter it, and the #arms_race resumes through the budget process rather than through any deliberate decision. This is why the loss of #verification, rather than the lapse of numerical limits as such, is the development that most worries specialists today. A limit without inspection is a statement of intent. An inspection regime without a limit still generates the shared factual picture that keeps estimates narrow. 11. Legacy and the Present Crisis 11.1 The chain of successor agreements START II, signed in January 1993, would have banned multiple-warhead land-based missiles altogether, addressing the core stability problem head on. It was ratified by both legislatures in different forms but never entered into force, and Russia set it aside in 2002 after the United States withdrew from the treaty limiting missile defences. The episode illustrates a general rule: offensive limits and defensive systems are politically linked, whatever the legal texts say. The 2002 Moscow Treaty reduced operationally deployed strategic warheads to a band of 1,700 to 2,200, in a document of remarkable brevity. It contained no #verification provisions of its own. It could function only because the machinery of #START_I was still operating. When that machinery expired in December 2009, the gap became urgent. #New_START, signed in 2010 and in force from 2011, was in part an attempt to rescue the #verification function. It set lower ceilings, at 1,550 deployed warheads and 700 deployed delivery vehicles, with 800 deployed and non-deployed launchers, and it simplified the inspection regime, using fewer inspection types and a streamlined counting system. It also fixed one of the older treaty's most obvious fictions by counting each deployed heavy bomber as one warhead, a simplification that was transparent about being a simplification. 11.2 Erosion and expiry The regime then unravelled in stages. On-site inspections under #New_START stopped in 2020 because of the pandemic and never resumed. In February 2023 Russia suspended its participation, while stating that it would continue to observe the numerical limits. Data exchanges and meetings of the consultative commission ceased. The treaty could not be extended a second time, because its text permitted only one extension, which had been used in 2021. On 5 February 2026 it expired. For the first time in more than fifty years, no treaty limits the strategic nuclear forces of the United States and the Russian Federation. In September 2025 the Russian president proposed that both sides continue to observe the central limits for one year beyond expiry, without #verification. No formal agreement was concluded before the deadline. Both governments have since expressed interest in a successor arrangement, and the American position has emphasised the inclusion of China, which has declined to join numerical limits on the grounds that its arsenal is far smaller. 11.3 What the 1991 case says about 2026 Three implications follow from the analysis above. First, an informal understanding to observe old limits without inspection is worth something, but much less than it appears. It preserves a political norm and it may slow force growth. It does not preserve the shared factual picture, and it is that picture, not the number itself, that restrains worst-case planning. The historical precedent of observing an unratified agreement in the early 1980s is often cited as encouraging, but that arrangement existed while negotiations continued and a successor was expected. Today no negotiations are under way. Second, the machinery of #verification decays. Inspectors retire, laboratories close, procedures are forgotten, and the bureaucratic units that manage compliance are disbanded or defunded. The regime built between 1987 and 1994 took years of technical work and cost a great deal of money. If it is allowed to lapse entirely, rebuilding it will not be a matter of signing a document. This is the strongest practical argument for preserving some inspection activity even in the absence of numerical limits (Kristensen, Korda, and Rogers 2022). Third, the bilateral model is necessary but no longer sufficient. Any future framework will have to grapple with a third major arsenal, with non-strategic weapons, with novel delivery systems such as long-range torpedoes and nuclear-powered cruise missiles that fit none of the old categories, with missile defences, with long-range conventional strike, and with the integration of artificial intelligence into warning and command systems (Johnson 2023; Cunningham 2025; Mount 2025). The 1991 treaty cannot be copied. Its components, meaning declared data, agreed definitions, negotiated counting rules, elimination protocols, and layered inspection, can be. 12. Discussion: Lessons for Students of International Security Several general propositions can be drawn from this case, and they are the reason it appears in every syllabus on arms control. Definitions are power. More negotiating time went into defining a heavy bomber, a new type of missile, and a deployed launcher than into agreeing the headline numbers. Whoever controls the definitions controls what the treaty actually limits. Students who skip the annexes miss the treaty. Verification is a design problem, not a moral one. The question is never whether to trust an adversary. It is what an adversary would have to do in order to cheat in a militarily significant way, and what observable traces that behaviour would leave. Design the inspection to catch the trace. Reductions and stability are different goals. A treaty can cut numbers while making a crisis more dangerous, or leave numbers high while making one safer. #START_I pursued both, and its structure of sublimits shows how the two objectives were traded against each other. Treaties outlive the states that sign them, but only if someone chooses to save them. The #Lisbon_Protocol was not inevitable. It was a diplomatic construction that could easily have failed, in which case four nuclear-armed successor states might have emerged from the Soviet collapse. Assurances are not guarantees. The #Budapest_Memorandum offers the clearest lesson available on the difference between a political commitment and an enforceable obligation, and the consequences of blurring the two are now visible in the calculations of states around the world. Institutions consume the machinery built by their predecessors. The 2002 treaty had no #verification of its own and did not need any, until suddenly it did. Machinery built for one agreement becomes infrastructure for the next, and infrastructure that is not maintained disappears. 13. Limitations of This Study and Directions for Further Research This article is based on public documents and secondary scholarship. It cannot speak to internal deliberations that remain classified, particularly on the Russian side, and its treatment of the negotiating endgame therefore relies on published accounts whose authors were participants and who have their own perspectives. Three research agendas seem especially productive. The first is comparative. The elimination procedures of the 1991 treaty, the South African dismantlement, and the partial approaches attempted elsewhere could be compared systematically to build a general theory of what makes reduction stick (Pretorius 2024; Elbahtimy 2024). The second is technical and legal. New #verification methods that allow a warhead to be authenticated without revealing its design are under active development, and their maturity will determine whether any future treaty can move from counting launchers to counting weapons (Cotta-Ramusino et al. 2023; Herzog 2025). The third is political and quantitative. The relationship between #transparency, threat perception, and force procurement is asserted constantly in this literature and tested rarely. The end of the New START regime provides, unfortunately, a natural experiment in what happens to force planning when shared information disappears. 14. Conclusion The Strategic Arms Reduction Treaty of 1991 was the high point of a particular way of managing nuclear danger: slow, technical, legalistic, and grounded in the belief that adversaries can build a common description of reality even when they agree on nothing else. It required both superpowers to destroy weapons they had spent decades and enormous sums acquiring. It survived the disappearance of one of its signatories. It converted three post-Soviet states into non-nuclear-weapon states. It gave the world fifteen years in which the two largest arsenals were counted, inspected, and known. It also failed to do many things. It did not reduce stockpiles, only deployments. It ignored an entire class of shorter-range weapons. Its counting rules embedded convenient fictions. And it was built for a bipolar world that no longer exists. The correct assessment is therefore neither celebratory nor dismissive. #START_I was an engineering achievement in the management of #distrust. Its numbers are obsolete, but its methods are not, and no better method has been invented since. In July 2026, with the last bilateral cap expired and no successor in force, the treaty's real significance is that it stands as proof of concept. Verified reduction between hostile nuclear powers is not a theoretical possibility. It has been done, at scale, under worse conditions than most people remember. That it has been allowed to lapse is a political choice, not a technical inevitability, and political choices can be revisited. References Budjeryn, M. (2023). Inheriting the Bomb: The Collapse of the USSR and the Nuclear Disarmament of Ukraine. Baltimore: Johns Hopkins University Press. https://doi.org/10.1353/book.103315 Colbourn, S. (2022). Euromissiles: The Nuclear Weapons That Nearly Destroyed NATO. Ithaca, NY: Cornell University Press. Cotta-Ramusino, P., Maiani, L., Martellini, M., and Ragaini, R. (Eds.). (2023). Nuclear Risks and Arms Control: Problems and Progresses in the Time of Pandemics and War. Springer Proceedings in Physics 291. Cham: Springer. Cunningham, F. S. (2025). Under the Nuclear Shadow: China's Information-Age Weapons in International Security. Princeton, NJ: Princeton University Press. Elbahtimy, H. (2023). Approaching irreversibility in global nuclear politics. Journal for Peace and Nuclear Disarmament, 6(2), 199-217. https://doi.org/10.1080/25751654.2023.2295595 Elbahtimy, H. (2024). Developing empirical insights into irreversibility in nuclear arms control and disarmament. Journal for Peace and Nuclear Disarmament. https://doi.org/10.1080/25751654.2024.2371546 Fleck, D. (2023). Irreversibility in nuclear arms control and disarmament law? Journal of Conflict and Security Law, 28(3), 423-466. https://doi.org/10.1093/jcsl/krad013 Herzog, S. (2025). Dual-track nuclear disarmament: Institutional design of the Treaty on the Prohibition of Nuclear Weapons. The Nonproliferation Review, 32(1-3). https://doi.org/10.1080/10736700.2025.2591997 Johnson, J. (2023). AI and the Bomb: Nuclear Strategy and Risk in the Digital Age. Oxford: Oxford University Press. Kristensen, H. M., Korda, M., Johns, E., and Knight, M. (2024). Russian nuclear weapons, 2024. Bulletin of the Atomic Scientists, 80(2), 118-145. https://doi.org/10.1080/00963402.2024.2314437 Kristensen, H. M., Korda, M., Johns, E., and Knight, M. (2025). United States nuclear weapons, 2025. Bulletin of the Atomic Scientists, 81(1), 53-79. https://doi.org/10.1080/00963402.2024.2441624 Kristensen, H. M., Korda, M., and Rogers, J. (2022). The long view: Strategic arms control after the New START Treaty. Bulletin of the Atomic Scientists, 78(6). https://doi.org/10.1080/00963402.2022.2133287 Mount, A. (2025). Conventional deterrence of nuclear use. International Security, 50(2), 95-129. https://doi.org/10.1162/ISEC.a.13 Narang, V., and Sagan, S. D. (Eds.). (2022). The Fragile Balance of Terror: Deterrence in the New Nuclear Age. Ithaca, NY: Cornell University Press. Plokhy, S. (2021). Nuclear Folly: A History of the Cuban Missile Crisis. New York: W. W. Norton. Pretorius, J. (2024). Staying the course: Lessons from South Africa for irreversibility of nuclear disarmament. Journal for Peace and Nuclear Disarmament, 7(1), 41-59. Ritchie, N. (2023). Irreversibility and nuclear disarmament: Unmaking nuclear weapon complexes. Journal for Peace and Nuclear Disarmament, 6(2), 218-243. Rodgers, J., and Williams, H. (2023). The irreversibility paradox: What makes for enduring arms control and disarmament. Journal for Peace and Nuclear Disarmament, 6(2), 244-262. https://doi.org/10.1080/25751654.2023.2292812 Roehrlich, E. (2022). Inspectors for Peace: A History of the International Atomic Energy Agency. Baltimore: Johns Hopkins University Press. Stockholm International Peace Research Institute. (2023). SIPRI Yearbook 2023: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Stockholm International Peace Research Institute. (2025). SIPRI Yearbook 2025: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Woolf, A. F. (2024). Irreversibility in nuclear arms control: Lessons from the US-Soviet/Russian arms control process. Journal for Peace and Nuclear Disarmament, 7(1), 27-40. https://doi.org/10.1080/25751654.2024.2359229 #Strategic_Arms_Reduction_Treaty #START_I_1991 #US_USSR_relations #nuclear_weapons_policy #arms_control_history #strategic_offensive_arms #nuclear_verification #Cold_War_diplomacy #post_Soviet_denuclearisation #Lisbon_Protocol_1992 #nuclear_nonproliferation #strategic_stability_studies #international_security #treaty_law #disarmament_studies

The Treaty on Conventional Armed Forces in Europe, signed in Paris on 19 November 1990, was the most ambitious conventional #arms_control_regime ever attempted. It placed equal, legally binding and verifiable ceilings on five categories of heavy military equipment across the whole area from the Atlantic Ocean to the Ural Mountains, and it obliged states to destroy or convert everything held above those ceilings. This article offers a structured account of the treaty for students of international security, international law and European politics. It traces the origins of the agreement in the failed Mutual and Balanced Force Reduction talks, examines the mandate negotiations of 1987 to 1989, and reconstructs the architecture of the treaty itself: the two groups of states parties, the five equipment categories, the nested zones, the flank rules, the sufficiency rule, the reduction timetable and the intrusive inspection system. The article then follows the treaty through its most successful phase, when more than fifty thousand pieces of heavy equipment were verifiably removed from European arsenals, and through its long decline, from the flank dispute of the mid-1990s and the unratified Adapted Treaty of 1999, to the Russian suspension of 2007, the collapse of data exchange in 2011, Russia's formal withdrawal on 7 November 2023 and the suspension of the treaty by allied states parties in the weeks and months that followed. The central argument is that the treaty failed not because its technical design was weak but because that design was tied to a political order that ceased to exist. The conclusion draws lessons for any future attempt to rebuild conventional #military_stability in Europe. Keywords: conventional arms control; European security; treaty verification; arms reduction; NATO; Russia; OSCE; regime decay 1. Introduction Few international agreements have been described so often as a cornerstone of #European_security, and few have collapsed so quietly. The Treaty on Conventional Armed Forces in Europe, usually shortened to the CFE Treaty, was signed by twenty-two states in Paris on 19 November 1990, at the same summit meeting that produced the Charter of Paris for a New Europe. It entered into force on 9 November 1992. Its aim was simple to state and extremely hard to achieve: to remove the capacity of either side in Europe to launch a #surprise_attack or a large-scale offensive operation, by cutting the heavy equipment needed for such operations down to equal and openly counted levels. For roughly a decade, the treaty worked. Tens of thousands of tanks, armoured vehicles, artillery pieces, aircraft and helicopters were cut up, crushed, blown apart or converted to civilian use, in front of foreign inspectors who were entitled to watch. Thousands of inspections took place on the territory of states that had, only a few years earlier, planned to fight each other. States exchanged detailed information every year about the location, type and number of their major weapons, down to the level of individual units and storage sites. The treaty made the military map of Europe visible to everyone who was party to it. Then it stopped working. The reasons were political rather than technical. The treaty had been designed to manage a confrontation between two alliances of roughly equal weight. Within two years of signature, one of those alliances no longer existed and one of its members, the Soviet Union, had broken into fifteen states. The treaty was patched, extended and reinterpreted. A wholly new version, the #Adapted_CFE_Treaty, was agreed in Istanbul in 1999, but it never entered into force because Western states linked their #ratification to Russian troop withdrawals from Georgia and Moldova, and Russia refused to accept that linkage. In 2007 Russia declared a "suspension" of its obligations. In 2011 most Western parties stopped exchanging data with Moscow. In 2015 Russia stopped attending the treaty's standing body in Vienna. On 7 November 2023 Russia completed its formal #withdrawal, and on the same day the North Atlantic Council announced that allied states parties would suspend the operation of the treaty for as long as necessary. Over the following months several other parties, including Greece, Moldova and Turkey, suspended implementation as well. What remains is a treaty that is technically still in force for a small number of states but that no longer regulates anything. This article is written for students who need a clear, accurate and reasonably complete account of the treaty: what it said, how it was verified, what it achieved, and why it died. The account is deliberately structured like a research paper. It sets out aims and questions, explains method and sources, builds a conceptual frame, presents the historical and legal material, analyses the causes of decline, and draws lessons. Simple language is used throughout, but nothing important is simplified away. Where figures are contested or where sources disagree, that is stated rather than hidden. The wider reason for studying a dead treaty is that the problem it tried to solve has come back. Europe in the middle of the 2020s contains the largest land war on the continent since 1945, rearmament programmes in almost every capital, and no functioning framework for limiting or even counting conventional forces. If a framework is ever rebuilt, it will be built by people who have understood why the last one failed. That is the practical justification for the present study. 2. Aims, Research Questions and Contribution The aim of this article is to provide an evidence-based, teachable analysis of the treaty across its whole life cycle, from negotiation to collapse, and to connect that analysis to the general theory of #cooperative_security. The study is organised around five research questions. First, what political and military conditions made a legally binding conventional agreement possible in 1989 and 1990, after fifteen years of failure in earlier talks? Second, how exactly was the treaty designed, and what does that design tell us about the assumptions its authors held about war, stability and trust? Third, how effective was the #verification system in practice, and what is the evidence for the treaty's achievements? Fourth, why did the treaty decay, and at what point did decay become irreversible? Fifth, what lessons can be drawn for any future conventional arms control effort in Europe? The contribution of the article is mainly synthetic and pedagogical rather than archival. It does not present new primary documents. It brings together the legal text, the reported implementation record and the recent scholarly literature on the decline of European arms control, and organises that material into a single narrative that a student can follow from beginning to end. In doing so, it argues for a specific interpretation, set out fully in section 13: that the treaty's greatest strength, its precise fit with the bloc-to-bloc structure of 1990, became its fatal weakness once that structure disappeared, and that the failure of adaptation was a failure of political bargaining rather than of technical imagination. 3. Method and Sources This is a qualitative, document-based study. It uses three kinds of material. The first is the legal and diplomatic record: the text of the treaty itself, together with its associated protocols and annexes, the 1992 Tashkent and Oslo documents that redistributed Soviet entitlements, the 1992 Concluding Act of the Negotiation on Personnel Strength known as #CFE_1A, the 1996 Flank Document, the 1999 Adapted Treaty, and the public statements issued by governments and by the North Atlantic Council when the treaty was suspended and abandoned. These documents establish what the parties were legally obliged to do and what they said they were doing. The second is the implementation record as reported by governments and by independent monitoring bodies. Figures on equipment destroyed, inspections conducted and data exchanged come from official reporting and from long-running research annuals. Such figures should be treated with a degree of caution. Different sources count different things over different periods, and totals reported for the treaty's whole life are naturally larger than totals reported for the formal forty-month reduction period alone. Where this article gives a number, it indicates the basis of the count. The third is the secondary literature: books and journal articles on arms control theory, on the end of the Cold War, on NATO enlargement and on the collapse of the European security order. Recent scholarship has been preferred, since the interpretation of the treaty's death has changed sharply since 2022. The method is straightforward process tracing. The article identifies the key decision points in the treaty's history, asks what alternatives were available at each point, and asks what the decision that was actually taken reveals about the interests and beliefs of the parties. This method cannot prove causation in a statistical sense. It can, however, show which explanations are consistent with the sequence of events and which are not. Two limitations should be stated at the start. First, archival material from Russian, American and NATO decision-making is unevenly available, and the account of motives, especially Russian motives, therefore rests partly on public statements that may not reflect private reasoning. Second, the article is written by an outsider to the negotiations and depends on the accuracy of published reporting. These limitations are revisited in section 15. 4. Conceptual Framework: What Conventional Arms Control Was Supposed to Do To evaluate the treaty, we need a clear idea of what such an agreement is for. Classical arms control theory, developed in the late 1950s and 1960s, rests on a counter-intuitive claim: that adversaries who cannot trust each other may still share an interest in limiting the means of war, because uncontrolled competition can produce outcomes that neither side wants. On this view, the purpose of arms control is not disarmament for its own sake. It is the reduction of the risk of war, the reduction of the damage if war occurs, and the reduction of the economic cost of preparing for war. Three ideas from that tradition matter here. The first is the #security_dilemma. Measures that one state takes for defence may look offensive to another. A large armoured force positioned close to a border is frightening whatever its owner's intentions, because intentions are invisible while capabilities are not. The response is usually a counter-build-up, and both sides end up less secure and poorer. Conventional arms control tries to break this cycle by making capabilities smaller, more distant and more visible. The second is #stability. Arms control theorists distinguish between the balance of forces and the stability of that balance. A situation is stable when neither side gains a decisive advantage by striking first. In the conventional context of Cold War Europe, instability came from the massed armoured and artillery formations of the Warsaw Treaty Organisation, which were thought capable of a fast, deep offensive with little warning. The design objective of any serious agreement was therefore to raise warning time and to remove the equipment needed for a short-notice offensive. The third is #transparency. Even where numbers cannot be reduced very far, knowing what the other side has, and where it is, reduces the risk of miscalculation. Transparency also converts a private military secret into a shared public fact, which makes sudden changes in posture politically costly. This is the logic behind #confidence_building_measures and behind the annual #information_exchange that the treaty required. To these three ideas the treaty added a fourth element that is more political than theoretical: #reciprocity. Every obligation was symmetrical in form. Both groups faced the same ceilings, the same reduction rules and the same inspection burdens, even though the actual cuts fell overwhelmingly on one side. This symmetry gave the agreement a claim to fairness that mattered enormously in domestic ratification debates. From a regime perspective, we may also ask what makes an arms control regime durable. Four conditions are usually proposed. There must be a shared understanding of the problem. There must be rules that are precise enough to be checked. There must be a mechanism for resolving disputes before they become crises. And there must be a continuing political relationship in which compliance is worth more to each party than defection. The treaty scored very well on the first three conditions and, in the end, catastrophically badly on the fourth. This framework guides the analysis in section 13. 5. Historical Background: From MBFR to the Mandate Conventional arms control in Europe did not begin in 1989. It began, and failed, in Vienna in 1973. The talks on Mutual and Balanced Force Reductions, known as #MBFR, ran for sixteen years between NATO and Warsaw Pact states. They covered a narrow zone in central Europe and they focused on troop numbers rather than equipment. They produced nothing. The reason was a dispute about data. The two sides could not agree on how many soldiers the Eastern side actually had in the area, and without agreed baseline numbers there was no way to define a reduction. Each side accused the other of manipulating counts. The talks became a byword for diplomatic futility. Alongside MBFR, a second and more productive track developed within the Conference on Security and Co-operation in Europe. The #Helsinki_Final_Act of 1975 committed participating states to give advance notice of large military exercises. The Stockholm Document of 1986 went further, adding compulsory notification thresholds, observation of exercises and, for the first time in Europe, mandatory on-site inspection with no right of refusal. The Stockholm agreement is the direct ancestor of the treaty's verification regime, and it proved something important: that the Soviet Union could accept foreign inspectors on its soil. Three changes then made a comprehensive equipment treaty possible. The first was the arrival of Mikhail Gorbachev and the doctrine of "reasonable sufficiency", which accepted that the Soviet Union's enormous conventional superiority was politically counterproductive and economically ruinous. In December 1988 Gorbachev announced unilateral cuts of half a million troops and thousands of tanks, and stated that Soviet forces in Eastern Europe would be restructured into a clearly defensive posture. This was a signal that Moscow was prepared to accept asymmetric reductions, which had been the sticking point for fifteen years. The second was the precedent of the Intermediate-Range Nuclear Forces Treaty of 1987, which eliminated an entire class of missiles and established an unprecedented inspection system, including permanent portal monitoring. It demonstrated that asymmetric cuts and intrusive verification were negotiable. The third was a change of frame. Instead of counting soldiers, negotiators decided to count equipment. Equipment can be seen, photographed, counted and destroyed. It is far harder to hide a tank than a soldier, and far easier to verify that a tank has been cut in half than that a division has been disbanded. This shift from personnel to hardware is the single most important technical decision in the whole story, and it is what allowed the negotiations to escape the data trap that had killed MBFR. In January 1989 the participating states of the CSCE concluded the Vienna Concluding Document, which set out a #Mandate_for_Negotiation on conventional armed forces. The mandate was limited to the twenty-three members of the two alliances, and it defined the area of application as the whole of Europe from the Atlantic to the Urals. The mandate also stated the objectives that the treaty would later restate: the elimination of disparities harmful to stability, and the elimination of the capability for surprise attack and for large-scale offensive action. 6. The Negotiation, 1989 to 1990 Formal talks opened in Vienna in March 1989 and concluded in November 1990. Twenty months is astonishingly fast for an agreement of this complexity, and the speed is itself an object lesson: arms control moves when the political leadership wants it to move. Several features of the negotiation deserve attention. The first is the choice of categories. The parties agreed to limit five types of equipment, later collectively named #treaty_limited_equipment: #battle_tanks, #armoured_combat_vehicles, artillery of 100 millimetres calibre and above, #combat_aircraft, and #attack_helicopters. These were selected because they are the tools of offensive ground operations and of the air support required for them. Naval forces were excluded, at Soviet insistence unsuccessfully resisted by the West, and this exclusion was one of the compromises that made agreement possible. Ground-based air defence, most trucks, and light equipment were also excluded. The second is definition. A surprising share of the negotiation was spent on defining what a tank actually is, and where the boundary lies between an armoured personnel carrier, an armoured infantry fighting vehicle and a heavy armament combat vehicle. These arguments were not pedantic. A definition that is even slightly loose creates a category of equipment that escapes the limits, and the treaty's authors knew that any loophole would eventually be exploited. The resulting Protocol on Existing Types, with its detailed lists of named vehicles and aircraft, is one of the most technically precise documents in the history of arms control. The third is the collapse of the political ground under the negotiators' feet. The talks began while the Berlin Wall still stood and the Warsaw Treaty Organisation still functioned. Within nine months the Wall was gone, Germany was moving towards unification, and the Eastern alliance was disintegrating from the inside. The negotiators responded by pressing forward rather than reopening the design. The treaty was deliberately signed before the Eastern bloc dissolved formally, because everyone understood that a treaty between two groups could be signed only while two groups existed. This decision secured the agreement, but it also locked in a structure that was already becoming obsolete on the day of signature. That paradox sits at the centre of the story. A fourth feature is the connection to German unification. The Treaty on the Final Settlement with Respect to Germany, signed in September 1990, contained ceilings on the size of the future unified German armed forces and tied certain of those commitments to the conclusion of the conventional treaty. The two agreements were parts of a single political settlement in which the Soviet Union accepted a unified Germany inside NATO in exchange for, among other things, a legally binding and verifiable limit on Western conventional power. The text was approved by the twenty-two negotiating states in Vienna on 15 November 1990 and signed in Paris four days later. 7. The Architecture of the Treaty 7.1 Parties and groups The treaty was signed by twenty-two states: the sixteen members of #NATO at that time and the six European members of the #Warsaw_Pact. Its obligations were expressed not for individual states but for two "groups of states parties". Each group as a whole was bound by a ceiling, and it was left to the members of each group to divide that ceiling among themselves through internal agreements. This group logic is essential to understanding both the treaty's elegance and its later fragility. When the Soviet Union dissolved at the end of 1991, its obligations had to be divided among the successor states lying inside the area of application. This was done through the #Tashkent_Agreement of 15 May 1992, in which eight successor states, Russia, Ukraine, Belarus, Moldova, Georgia, Armenia, Azerbaijan and Kazakhstan, accepted shares of the former Soviet entitlements and liabilities. A further conference in Oslo in June 1992 adjusted the treaty to reflect the new membership. Czechoslovakia later split into two parties. The result was a treaty with thirty states parties, which is the figure normally quoted for its mature period. 7.2 The area of application The treaty applied to land territory from the Atlantic Ocean to the Ural Mountains, a formula usually abbreviated as #Atlantic_to_the_Urals. This included the whole of European Russia and the European parts of the other successor states, and it excluded the United States and Canada, which were parties but whose home territory lay outside the zone. Equipment brought into the zone by outside parties counted against their entitlements, so American armour stationed in Germany was covered while American armour in Texas was not. 7.3 The five categories and the ceilings Each group was permitted a maximum holding in each category. The agreed figures were as follows. Category Ceiling for each group Sub-ceiling in active units Battle tanks 20,000 16,500 Armoured combat vehicles 30,000 27,300 Artillery, 100 mm and above 20,000 17,000 Combat aircraft 6,800 not applicable Attack helicopters 2,000 not applicable Two refinements deserve note. Within the armoured combat vehicle ceiling, no more than 18,000 could be armoured infantry fighting vehicles or heavy armament combat vehicles, and of those no more than 1,500 could be heavy armament combat vehicles. This prevented a group from filling its allowance with the most powerful vehicles. Equipment above the active-unit sub-ceilings had to be placed in designated permanent storage sites, where it could be inspected and where it could not be used quickly. 7.4 The sufficiency rule Article VI contained what is usually called the #sufficiency_rule. No single state party could hold more than roughly one third of the total equipment in the area of application in any category. In practice this meant a national maximum of about 13,300 battle tanks, 20,000 armoured combat vehicles, 13,700 artillery pieces, 5,150 combat aircraft and 1,500 attack helicopters. The rule was aimed at the Soviet Union, and after 1991 at Russia, but it applied to everyone. Its purpose was to prevent any one state from dominating the continent even if its allies agreed to give it their shares. 7.5 Zones and the flank rules The treaty did not only limit how much equipment could exist. It limited where it could be. Articles IV and V divided the area of application into a set of nested zones. The innermost zone covered central Europe, where the two alliances faced each other most directly, and it carried the tightest limits. Moving outward, the permitted holdings increased. Because the zones were nested, equipment counted simultaneously against the limit of the zone it was in and the limits of all wider zones containing it. The effect was to prevent the massing of forces near the central front while still allowing states to keep equipment further back. Article V established the special regime known as the #flank_limits. The flank zone covered the northern and southern extremities of the area, including Norway, Iceland, Greece, Turkey, Bulgaria, Romania and, on the Eastern side, the Leningrad and North Caucasus military districts of the Soviet Union together with Moldova, Georgia, Armenia and Azerbaijan. Holdings in this zone were capped separately. The rule made sense in 1990: it stopped the Soviet Union from moving armour out of central Europe and parking it on NATO's northern and southern edges, which was a serious Norwegian and Turkish concern. After 1991 the flank rule became the treaty's most explosive provision. The North Caucasus was now the front line of Russia's internal wars, and Moscow argued that a rule written to protect Norway and Turkey from Soviet invasion was preventing it from fighting insurgents on its own territory. The flank dispute, examined in section 12, was the first crack in the structure. 7.6 Reductions States holding equipment above their entitlements were required to eliminate the excess. The reduction liability was enormous, and it fell almost entirely on the Eastern side, since NATO's holdings were already close to or below the new ceilings in several categories while Soviet holdings were far above them. The reduction period ran for forty months from entry into force, ending in November 1995, and was divided into three phases. At least twenty-five per cent of a state's total liability had to be eliminated within sixteen months, at least sixty per cent within twenty-eight months, and one hundred per cent within forty months. This phasing mattered, because it meant that a party could not simply promise cuts and then delay them until the political weather changed. The methods of #destruction_of_equipment were specified in detail in a protocol. Tanks could be destroyed by severing, by explosive demolition, by smashing, or by use as targets. Armoured vehicles and artillery had comparable methods. A limited number of items could instead be converted to non-military purposes, such as bulldozers, fire-fighting vehicles or cranes, or could be placed on static display, used as ground instructional airframes, or in the case of certain trainer aircraft reclassified. Every reduction had to be carried out at a declared reduction site and had to be open to inspection, and states had to notify in advance which items they intended to reduce and when. The important point for students is that the treaty did not accept promises. It accepted only wreckage that a foreign inspector had been allowed to look at. 7.7 Duration and exit The treaty was of unlimited duration. #Article_XIX provided a right of withdrawal if a party decided that extraordinary events related to the subject matter of the treaty had jeopardised its supreme interests, subject to one hundred and fifty days' notice and a statement of the extraordinary events concerned. Notably, the text contained no provision at all for #suspension, a gap that became legally significant in 2007 and again in 2023. 8. The Verification and Information Regime If the ceilings were the treaty's skeleton, verification was its nervous system. The system had two halves: information and inspection. 8.1 The information regime Every party had to provide a detailed annual exchange of information on its forces in the area of application. The exchange covered the structure of its land, air and air-defence forces, the location and holdings of every formation and unit down to a specified level, the location of every declared site holding treaty-limited equipment, and the numbers of each item type at each site. Information was exchanged in December and took effect on the first of January. Additional notifications were required when units were created, disbanded or moved, when equipment entered or left the area, and when equipment was scheduled for reduction. The cumulative effect was remarkable. For roughly two decades, states that had spent forty years trying to guess each other's order of battle simply had it delivered to them in a standard format, annually, in writing. The information regime is arguably the treaty's most under-appreciated achievement, because it changed the baseline conditions of European intelligence and planning. 8.2 Inspections Information alone is worth little without the right to check it. The treaty gave each party the right to conduct #on_site_inspection on the territory of other parties, and imposed on each party a duty to accept a quota of inspections calculated as a share of its declared objects of verification. An object of verification was, roughly, a unit or a site holding treaty-limited equipment. The more objects a state declared, the more inspections it had to accept. The quota was heavier during the baseline validation and residual level validation periods, when the accuracy of the declared data mattered most, and lighter during routine years. There were three principal inspection types. Declared site inspections allowed a team to visit a site named in the data exchange and to count what was there against what had been declared. The inspected party could not refuse. #Challenge_inspections, formally inspections within specified areas, allowed a team to demand access to an area that had not been declared, on suspicion that undeclared equipment was being held there. Refusal was possible but carried a heavy political cost, since the inspected state then had to explain why. Reduction inspections allowed a team to observe the destruction or conversion of equipment at a reduction site, and to confirm that the items destroyed were the items declared. Inspection teams had short notice periods, guaranteed rights of access within the site, and the right to use specified equipment such as cameras and measuring instruments. Aerial observation was not part of the treaty, but the #Open_Skies_Treaty of 1992 later added an overflight regime that complemented it. 8.3 What verification could and could not do Verification was strong on stock and weak on flow. It could establish with high confidence how many tanks were sitting in a declared depot on a given day. It was less able to track equipment that moved rapidly, and it could do nothing at all about equipment held outside the area of application, which is why Russian holdings east of the Urals were a persistent Western complaint and why the movement of thousands of vehicles beyond the Urals before entry into force was so controversial. Verification also depended on cooperation. It could detect and document a violation, but it could not compel a remedy. Once a state simply stopped accepting inspectors, as Russia did after 2007, the entire system had nothing left to work with. This is the fundamental limit of #compliance monitoring in a treaty without an enforcement mechanism. It converts secret cheating into open defiance, which is valuable, but it cannot convert open defiance into obedience. 9. Institutional Machinery: The Joint Consultative Group The treaty established a standing body in Vienna, the #Joint_Consultative_Group, in which all parties were represented and which met regularly, in practice weekly, throughout the treaty's active life. Its functions were to address questions of compliance, to resolve ambiguities in interpretation, to consider disputes, to consider measures to enhance the treaty's effectiveness, and to receive and circulate agreed information. The group worked by consensus. This gave every party, including the smallest, a veto, and it meant that no interpretation could be imposed on an unwilling state. In its good years the group was an effective technical clearing house, quietly settling counting disputes and definitional problems that would otherwise have escalated. Its decline is a useful indicator of the treaty's health. Once Russia suspended participation in 2007, the group could no longer address the most important compliance questions, and once Russia stopped attending altogether in March 2015, the body became a forum in which the states that were still implementing the treaty talked to each other about a regime that no longer covered the state it had been designed to constrain. A general lesson follows. Consensus institutions are excellent at managing disputes between parties that want the dispute managed. They are helpless when a major party decides that the dispute is the point. 10. Implementation and Results, 1992 to 2007 The treaty's operational achievements were real and should not be diminished by its eventual failure. During the forty-month reduction period ending in November 1995, states parties verifiably eliminated well over fifty thousand items of treaty-limited equipment. Counts that cover the treaty's full working life, including reductions carried out after the formal reduction period and the destruction associated with later force restructuring, run considerably higher, and figures above seventy thousand items are commonly cited. The variation in reported totals reflects different counting periods rather than disagreement about the underlying facts, and students quoting a number should always state which period it refers to. Thousands of on-site inspections were conducted. Inspectors from formerly hostile states visited each other's bases, storage depots and reduction sites, and the routine of inspection created a professional network of officers across Europe who knew each other personally. This human by-product of verification is rarely captured in statistics but was frequently noted by participants as one of the treaty's most valuable effects. The pattern of #arms_reductions also changed the shape of European forces in ways that outlasted the treaty. Because equipment above the active-unit sub-ceilings had to be stored rather than deployed, and because the zonal limits pushed heavy forces away from the old central front, the massed armoured groupings that had defined European military planning since the 1940s were dismantled. Western European states, freed from the requirement to hold large armoured reserves, cut far below their entitlements, and by the 2000s most NATO members in Western Europe held only a fraction of what the treaty would have allowed. The treaty did not cause this so much as legitimise and lock in a process driven by the end of the Cold War and by budgetary pressure, but it gave that process a legal shape and made it mutually visible. Three qualifications must be entered against this record. First, the reductions were made possible by a political transformation that had already occurred. The treaty codified a shift in the balance of power rather than creating one. It is much easier to obtain deep cuts from a state that has already decided its previous posture was unaffordable. Second, compliance was not perfect. Disputes arose over equipment moved east of the Urals before the treaty took effect, over unaccounted-for equipment in the Caucasus, over Russian forces in Moldova and Georgia, and over holdings in the flank zone that Western states judged to exceed the limits. Some of these disputes were resolved in the Joint Consultative Group. Others festered. Third, the treaty's coverage was frozen. New NATO members that joined the alliance after 1990 but had not been parties in 1990, most importantly Estonia, Latvia, Lithuania and Slovenia, were not covered at all, because the original text made no provision for accession. This became one of Russia's central and, on its own terms, entirely coherent complaints: the treaty constrained Russian deployments while leaving a strip of NATO territory on Russia's border formally unlimited. 11. The Companion Agreements The treaty was never meant to stand alone. Three related instruments formed the wider architecture. The first was the Concluding Act of the Negotiation on Personnel Strength, signed in Helsinki in July 1992 and known as CFE-1A. This was politically rather than legally binding, and it set national limits on the personnel strength of conventional forces in the area of application. It was the closest the parties ever came to solving the problem that had defeated MBFR, and it worked only because it was attached to a hardware treaty that could be verified. The second was the #Vienna_Document on confidence and security building measures, first adopted in 1990 and revised several times thereafter, most recently in 2011. The Vienna Document is an #OSCE political commitment rather than a treaty. It provides for annual exchanges of military information, prior notification and observation of major exercises, evaluation visits and inspections. Its thresholds are high and its inspection quotas are low compared with the treaty, so it is a much weaker instrument. It is nevertheless significant that it is the only element of the conventional architecture still standing, and that repeated attempts to modernise it since 2011 have been blocked. The third was the Open Skies Treaty, signed in 1992 and in force from 2002, which permitted unarmed observation flights over the entire territory of participating states. It was designed to complement ground inspection with aerial imagery. The United States withdrew in 2020 and Russia in 2021, which removed the aerial layer from the architecture before the ground layer collapsed. Taken together, these instruments show that the designers of the post-Cold War order thought in systems: limits, personnel caps, exercise notification and aerial observation were meant to reinforce one another. The system also failed as a system. Each element that fell weakened the case for the others, and by 2024 only the weakest element remained. 12. Erosion and Collapse The decline of the treaty is best understood as a chain of decisions, each of which made the next one more likely. 12.1 The flank dispute, 1993 to 1996 Almost as soon as the treaty entered into force, Russia asked for relief from the flank limits. Its argument had force. The flank rule had been written to constrain Soviet forces facing NATO, and the Soviet Union no longer existed. The North Caucasus, formerly a rear area, was now the site of a large internal war. Western states were reluctant. Norway and Turkey regarded the flank rule as the provision that most directly protected them. But a compromise was reached at the First Review Conference in Vienna in 1996. The Flank Document redrew the boundaries of the flank zone, moving certain Russian and Ukrainian territories out of it, and adjusted the permitted levels. It entered into force in 1997. The episode is usually described as a success of adaptation, and in narrow terms it was. But it set two precedents. It established that Russia would treat the treaty's geographic provisions as negotiable when they became inconvenient, and it established that Western states would eventually accommodate. Both lessons were learned, and both were later applied in circumstances where they did not produce agreement. 12.2 NATO enlargement and the Adapted Treaty, 1997 to 1999 The deeper problem was structural. A treaty built on two groups of comparable weight cannot function when one group grows and the other disappears. When Poland, Hungary and the Czech Republic joined NATO in 1999, they moved their entitlements from one group to the other. On paper, the balance shifted decisively. Russia argued, with obvious justification in arithmetic if not in threat assessment, that the group ceilings had become meaningless. The response was the Adapted Treaty, agreed at the OSCE summit in Istanbul in November 1999. It was a genuine and thoughtful redesign. It abolished the group structure entirely and replaced it with two new kinds of limit: #national_ceilings, which capped what each state could own, and #territorial_ceilings, which capped how much equipment could be present on each state's territory, including equipment belonging to other states. It provided for accession by states that had not been original parties, which would have brought the Baltic states inside the regime. It kept and in some respects strengthened the verification system. The Adapted Treaty was, in design terms, close to what the situation required. It never entered into force. 12.3 The Istanbul linkage At the same summit, Russia made a set of political commitments, generally known as the #Istanbul_Commitments, including the withdrawal of its forces from Georgia and the withdrawal of its forces and ammunition from the Transnistrian region of Moldova. NATO states subsequently took the position that they would ratify the Adapted Treaty only once these commitments had been fulfilled. Russia rejected the linkage as illegitimate, arguing that its Istanbul undertakings were separate political commitments with no legal connection to the Adapted Treaty, and that Western states were using them as a pretext. Western states argued that ratifying a treaty that would legitimise territorial ceilings while Russian forces remained on the territory of two other parties without their consent would reward the very behaviour the treaty existed to prevent, and that #host_nation_consent was a foundational principle of the regime. Both positions were internally coherent. Neither side moved. Only Russia, Belarus, Kazakhstan and Ukraine ratified the Adapted Treaty. The deadlock lasted eight years, during which the original treaty continued to operate while everyone knew that its replacement was frozen. This period, more than any single dramatic event, is where the regime was lost. A treaty that all parties agree is obsolete, and whose agreed replacement all parties agree is stuck, cannot survive indefinitely. 12.4 Suspension, 2007 In April 2007 President Putin announced a moratorium on Russian implementation. An extraordinary conference of states parties in Vienna in June failed to resolve the dispute. In July Putin signed a decree suspending Russian participation, and the suspension took effect on 12 December 2007. Russia set out conditions for return, including ratification of the Adapted Treaty by NATO members, accession by the four alliance members outside the regime, lower Western entitlements, and the removal of flank constraints on Russia. The legal position was awkward. As noted, the treaty contained no provision for suspension. Russia was therefore claiming a right that the text did not grant, on the basis of a general principle of international law relating to a fundamental change of circumstances. Western states rejected the claim but chose not to declare Russia in material breach, and continued to implement the treaty among themselves and to provide data to Russia for a further four years. This restraint was intended to keep the door open. It also allowed an asymmetric situation to persist in which one party enjoyed the transparency benefits of the regime without paying its costs. 12.5 The end of data exchange, 2011 On 22 November 2011 the United States and most other NATO parties announced that they would cease implementing treaty obligations with respect to Russia, while continuing to implement them with all other parties. Data would no longer be provided to Moscow and Russian inspection requests would no longer be accepted. This was a countermeasure, proportionate and reversible in principle. Its practical effect was limited, and the reason is instructive. Belarus remained a party and continued to receive treaty data, and it was widely assumed among allies that Belarus shared what it received with Russia. A transparency regime with a leak in it is not a transparency regime. The 2011 decision therefore imposed few costs on Russia while formalising the split. 12.6 Departure from the Joint Consultative Group, 2015 On 11 March 2015, in the year following the annexation of Crimea and the outbreak of war in eastern Ukraine, Russia announced that it would stop attending meetings of the Joint Consultative Group. Russian officials described the treaty as an anachronism that no longer corresponded to reality. This step removed the last channel in which the parties discussed the treaty with each other. From 2015 the regime was, in substance, over, although it took another eight years for the paperwork to catch up. 12.7 Withdrawal and suspension, 2023 and 2024 Russia initiated formal withdrawal procedures in mid-2023, following legislation denouncing the treaty, and the withdrawal was completed on 7 November 2023. In its statement, Moscow linked the decision to NATO enlargement and to what it described as the impossibility of maintaining outdated agreements. Russia stated that its withdrawal also annulled the associated agreements on flank holdings and on personnel strength. On the same day the North Atlantic Council issued a statement condemning the withdrawal and announcing that allied states parties intended to suspend the operation of the treaty for as long as necessary, on the ground that a situation in which allies abided by the treaty while Russia did not would be unsustainable. The United States confirmed that its suspension would take effect on 7 December 2023, and framed the step as a response to a fundamental change of circumstances arising from Russia's withdrawal combined with its war against Ukraine, itself a state party. The unravelling continued. Greece suspended implementation in February 2024, Moldova in March 2024, and Turkey suspended implementation with effect from 8 April 2024, while stressing that it had not withdrawn and that its decision was reversible. Turkish officials made the essential point plainly: with the principal object of the regime outside it and most other parties suspended, continued implementation had lost its meaning. The treaty has therefore not been formally terminated. It has been abandoned. It remains in force on paper for a small number of parties, none of which is implementing it in any meaningful way. Legally, this state of affairs is reversible. Politically, it is not, at least while the war in Ukraine continues. 13. Discussion: Why It Worked, and Why It Died 13.1 The sources of success Four factors explain the treaty's early achievements. The first is political will at the highest level. Soviet leaders had already decided that their conventional posture was unaffordable and counterproductive. Western leaders had decided that locking in Soviet reductions was worth accepting constraints on themselves. When both sides want an agreement, technical obstacles turn out to be soluble. The second is the shift from personnel to equipment. Counting hardware made #verification possible, which made deep cuts credible, which made ratification politically defensible. The whole edifice rests on this one design choice. The third is symmetry of form. Every party carried the same categories of obligation. The asymmetry of outcome was accepted because the rules themselves treated everyone alike. Legitimacy in arms control comes less from equality of outcome than from equality of rule, and the treaty understood this. The fourth is intrusiveness. Because inspection was frequent, mandatory and detailed, cheating on a scale large enough to matter militarily was extremely difficult. The regime did not require trust. It manufactured a substitute for trust, which is a far more reliable commodity. 13.2 The sources of failure The failure had one deep cause and several proximate ones. The deep cause was that the treaty's structure was welded to a political situation that vanished. The two-group architecture presupposed two alliances of comparable weight, a shared central front, and a stable membership on both sides. Within twenty-four months of signature, one alliance had dissolved, one superpower had broken up, and the front line had moved a thousand kilometres east. Everything that followed was an attempt to fit a 1990 instrument to a world it had not been designed for. The proximate causes follow from this. The first is the failure of adaptation. The Adapted Treaty of 1999 was a serious and technically sound answer to the structural problem. It died in the ratification stage because it became a bargaining chip in a separate dispute about Georgia and Moldova. Whether one regards that #linkage_politics as principled or as opportunistic depends on one's view of the underlying dispute, but its effect is not in doubt: it destroyed the only agreed solution to the treaty's structural flaw. The lesson is not that linkage is always wrong, since a treaty legitimising the presence of foreign troops without host consent would have created problems of its own. The lesson is that a linkage which cannot be resolved will eventually consume the thing to which it is linked. The second is divergent threat perception. From the mid-1990s onward, Western and Russian assessments of what the treaty was for diverged completely. Western states saw a regime for locking in transparency and preventing destabilising concentrations, whose limits were largely irrelevant because Western forces had shrunk far below them. Russia saw a regime that constrained Russian deployments on Russian territory while an expanding alliance faced no equivalent constraint and while its newest members were not covered at all. Both readings were accurate descriptions of the same document. That is precisely the problem: the treaty had become a mirror in which each side saw a different picture. The third is the erosion of the wider relationship. Arms control regimes are not self-supporting. They rest on a political relationship in which each side judges continued cooperation to be worth more than the freedom of action it gives up. From Kosovo in 1999 through the enlargement rounds, the 2008 war in Georgia, the annexation of Crimea in 2014 and the full-scale invasion of Ukraine in 2022, that judgement moved steadily in one direction. By 2023 the treaty was not the cause of anything. It was a piece of paper describing a relationship that had ended. The fourth is the absence of an enforcement mechanism. When Russia suspended in 2007, the other parties had no graduated response available except to stop implementing the treaty themselves, which they delayed for four years and which, when finally applied in 2011, was undermined by the Belarus channel. A regime that offers only two options, full compliance or mutual abandonment, will eventually be abandoned. 13.3 Was the treaty obsolete anyway? A serious counter-argument deserves attention. It holds that the treaty was becoming militarily irrelevant regardless of politics, because it counted the wrong things. The treaty limited platforms: tanks, armoured vehicles, tube artillery, manned aircraft and attack helicopters. It did not limit precision-guided munitions, cruise missiles, drones, electronic warfare systems, or the reconnaissance and command networks that make modern firepower effective. It did not limit naval forces at all. It said nothing about the reinforcement capacity of forces based outside the area of application, which for the United States is decisive. It contained no rules on private military companies, on cyber capabilities, or on the rapid mobilisation of stored civilian resources. The war in Ukraine has made the point vivid. A conflict in which cheap uncrewed systems destroy expensive armour, in which artillery ammunition production is a strategic variable more important than artillery tube counts, and in which long-range strike shapes the battlefield hundreds of kilometres behind the line, is not a conflict that a treaty counting tubes and turrets would have prevented or even much constrained. The counter-argument has real force, and any future regime will have to answer it. But it should not be pushed too far. The treaty's core function was not to predict future warfare. It was to remove the mass and proximity that make short-warning offensives possible, and to make forces visible to each other. Those functions remain valuable, and the war in Ukraine, which began with a large armoured invasion, is not evidence that heavy platforms have ceased to matter. It is evidence that they matter in different ways. The honest conclusion is that the treaty's counting rules were ageing, that its adaptation was overdue on military as well as political grounds, and that obsolescence contributed to its death without being the primary cause. 13.4 Interpretation Returning to the framework of section 4, the treaty satisfied three of the four conditions for a durable regime. It rested on a shared understanding of the problem, at least in 1990. It contained rules precise enough to check. It possessed a dispute mechanism, the Joint Consultative Group, which functioned well for a decade. What it lacked, in the end, was the fourth condition: a continuing political relationship in which compliance was worth more than defection. This suggests a general and rather uncomfortable conclusion about #cooperative_security. Arms control cannot substitute for a political settlement. It can stabilise a relationship, make it cheaper and safer, and give it a shared vocabulary. It cannot hold a relationship together against the will of the parties. Those who treat arms control as a cure for hostility will be disappointed. Those who treat it as a management tool for hostility that both sides wish to manage will find it extraordinarily useful. The treaty was the finest management tool ever built for European security, and it was discarded the moment one party decided that the relationship it managed was no longer one it wanted. 14. Lessons for Future Conventional Arms Control in Europe Any attempt to rebuild conventional #arms_control after the present war will confront the same problems in a harsher form. Six lessons follow from the treaty's history. First, build for change. The single greatest design error was structural rigidity. A future instrument must contain built-in procedures for accession by new states, for adjustment of ceilings, and for the incorporation of new categories of equipment without renegotiation of the whole text. Amendment procedures that require unanimous ratification are, in practice, procedures for paralysis. Second, separate the technical from the political where possible. The Istanbul linkage shows how a functioning technical regime can be destroyed by a dispute it was not designed to resolve. This does not mean that political conditions should never be attached, but it does mean that negotiators should ask, before attaching them, what happens to the regime if the condition is never met. Third, keep the transparency layer alive even when the limitation layer is dead. Limits require a settled balance of power. Transparency requires only a shared interest in avoiding accidental war, which even bitter enemies usually retain. The Vienna Document, weak as it is, is the surviving element precisely because it demands less. A modernised information and notification regime is a far more realistic near-term objective than new ceilings, and it should be pursued on its own terms rather than as a stepping stone that is abandoned if ceilings prove impossible. Fourth, count what matters now. A future regime will have to address uncrewed systems, long-range strike, ammunition stockpiles and reinforcement capacity, or it will regulate a museum. This is technically hard. Drones are cheap, numerous and dual-use, and a counting rule for them will be far more difficult to write than a counting rule for tanks. The difficulty is a reason to start work early, not a reason to avoid the problem. Fifth, expect asymmetry and design for it honestly. The original treaty achieved equal group ceilings because one side was willing to make almost all of the cuts. A future European settlement will involve a Russia that is militarily weaker in some respects and more heavily armed in others than the alliance facing it, and a NATO that has enlarged dramatically. Pretending to symmetry that does not exist will produce a text that no legislature ratifies. Regimes that acknowledge asymmetry and compensate for it with #confidence_building measures and #risk_reduction mechanisms have a better chance. Sixth, protect the human infrastructure. Verification regimes create communities of inspectors, negotiators and analysts. That community is now ageing and dispersing. Rebuilding it will take years, and no treaty text can substitute for people who know how to run an inspection and who have met their counterparts. States that are serious about future arms control should be training verification specialists now, in the absence of anything to verify. A final and more sober lesson concerns sequencing. The treaty was signed at a moment when the political settlement had already been reached and the arms control instrument merely recorded it. There is no comparable settlement in prospect in Europe today. It follows that the realistic near-term agenda is not a new treaty but the preservation of minimal #risk_reduction channels: notification of exercises, deconfliction mechanisms, and open lines of communication between military authorities. These are unglamorous. They are also what stands between a tense confrontation and an accidental one. 15. Limitations and Directions for Further Research This article has three limitations that a reader should keep in mind. First, it relies on public sources. The internal deliberations of the parties, particularly Russian decision-making in 2007 and 2023, are not accessible, and the account of motives given here is an inference from behaviour and public statements. Archival work, when it becomes possible, may revise it. Second, first-hand testimony is missing. The community of officials who negotiated and implemented the treaty is still available for interview, and oral history of the inspection regime would capture knowledge that exists in no document. This is an obvious and urgent research task. Third, the article treats the treaty largely as an inter-state phenomenon. It says little about domestic politics, about the role of legislatures in ratification battles, about defence industries, or about the bureaucratic interests of armed forces that disliked the constraints the treaty imposed on planning and exercises. Statements by allied governments after the 2023 suspension, noting with evident relief that suspension would remove restrictions affecting planning, deployments and exercises, suggest that these bureaucratic interests were real and that they deserve study. Four further research questions follow. What counting rules for uncrewed and precision systems would be technically verifiable? What can be learned from conventional arms control efforts outside Europe, where asymmetry is the norm? Does a regional or sub-regional approach, covering, for example, the Baltic or Black Sea areas, offer a more realistic path than a continental one? And what is the effect of a collapsed arms control regime on the risk of inadvertent #escalation_risk, which is now the central practical question for European security? 16. Conclusion The Treaty on Conventional Armed Forces in Europe was the most ambitious and the most successful conventional arms control agreement ever concluded. It set equal, legally binding ceilings on five categories of heavy equipment across a continent, obliged states to destroy tens of thousands of items under the eyes of foreign inspectors, and made the military posture of thirty states an open book to one another. For a decade it worked, and the Europe it helped create was safer, cheaper to defend and more predictable than the one it replaced. It failed because it was too well made for the world of 1990. Its architecture assumed two alliances of comparable weight facing each other across a central front. When that world dissolved, the treaty's precision became rigidity. The one serious attempt to redesign it, the Adapted Treaty of 1999, was technically sound and politically stillborn, killed by a linkage to a dispute it could not settle. From that point the regime lived on borrowed time: suspension in 2007, the end of data exchange in 2011, the abandonment of the consultative body in 2015, and finally Russian withdrawal in 2023, followed within months by suspension on the part of almost every remaining party. The deeper lesson is one that arms control practitioners have always known and that policy makers regularly forget. Treaties do not create trust. They manage the relations of states that have already decided, for reasons of their own, that management is worth the effort. When that decision is reversed, no amount of drafting skill will hold the structure up. The treaty's verification machinery was superb and it did not save the treaty, because verification answers the question of whether a state is complying, not the question of whether it wants to. Europe now has the largest conventional forces on its territory since the Cold War, an active war between two former states parties, and no functioning framework for limiting, notifying or counting anything of significance. The equipment is back. The rules are gone. The lesson of the treaty is not that such rules are futile, but that they must be rebuilt on political foundations that can bear their weight, and that the work of building those foundations cannot be delegated to the treaty itself. #CFE_Treaty #conventional_arms_control #European_security_architecture #treaty_verification #arms_control_history #NATO_Russia_relations #post_Cold_War_Europe #disarmament_studies #international_security #Paris_1990 #Istanbul_1999 #regime_decay #military_transparency #OSCE_arms_control #future_of_arms_control References Charap, S., & Priebe, M. (2023). Avoiding a long war: U.S. policy and the trajectory of the Russia-Ukraine conflict. RAND Corporation, Santa Monica, CA. Fix, L. (2021). Germany's role in European Russia policy: A new German power? Palgrave Macmillan, Cham. Freedman, L. (2022). Command: The politics of military operations from Korea to Ukraine. Oxford University Press, New York. Gotz, E., & Staun, J. (2022). Why Russia attacked Ukraine: Strategic culture and radicalized narratives. Contemporary Security Policy, 43(3), 482-497. Institute for Peace Research and Security Policy at the University of Hamburg (Ed.). (2021). OSCE Insights 2020. Nomos, Baden-Baden. Institute for Peace Research and Security Policy at the University of Hamburg (Ed.). (2022). OSCE Insights 2021. Nomos, Baden-Baden. Institute for Peace Research and Security Policy at the University of Hamburg (Ed.). (2023). OSCE Insights 2022. Nomos, Baden-Baden. Institute for Peace Research and Security Policy at the University of Hamburg (Ed.). (2024). OSCE Insights 2023. Nomos, Baden-Baden. International Institute for Strategic Studies. (2023). The Military Balance 2023. Routledge, London. International Institute for Strategic Studies. (2024). The Military Balance 2024. Routledge, London. International Institute for Strategic Studies. (2025). The Military Balance 2025. Routledge, London. Krepon, M. (2021). Winning and losing the nuclear peace: The rise, demise, and revival of arms control. Stanford University Press, Stanford, CA. Kuhn, U. (2020). The rise and fall of cooperative arms control in Europe. Nomos, Baden-Baden. Person, R., & McFaul, M. (2022). What Putin fears most. Journal of Democracy, 33(2), 18-27. Plokhy, S. (2023). The Russo-Ukrainian war: The return of history. W. W. Norton, New York. Radchenko, S. (2024). To run the world: The Kremlin's Cold War bid for global power. Cambridge University Press, Cambridge. Sarotte, M. E. (2021). Not one inch: America, Russia, and the making of post-Cold War stalemate. Yale University Press, New Haven, CT. Stockholm International Peace Research Institute. (2022). SIPRI yearbook 2022: Armaments, disarmament and international security. Oxford University Press, Oxford. Stockholm International Peace Research Institute. (2023). SIPRI yearbook 2023: Armaments, disarmament and international security. Oxford University Press, Oxford. Stockholm International Peace Research Institute. (2024). SIPRI yearbook 2024: Armaments, disarmament and international security. Oxford University Press, Oxford. Stockholm International Peace Research Institute. (2025). SIPRI yearbook 2025: Armaments, disarmament and international security. Oxford University Press, Oxford. Zubok, V. M. (2021). Collapse: The fall of the Soviet Union. Yale University Press, New Haven, CT.

The Convention on Prohibitions or Restrictions on the Use of Certain Conventional Weapons Which May Be Deemed to Be Excessively Injurious or to Have Indiscriminate Effects, adopted in Geneva on 10 October 1980 and in force since 2 December 1983, is the main multilateral treaty that applies two old rules of the law of war to specific categories of ordinary weapons. Those two rules are the ban on weapons that are #indiscriminate_effects by nature and the ban on weapons of a nature to cause #superfluous_injury or #unnecessary_suffering. This article studies the #CCW as a legal system, as an institution, and as a political forum. It explains the treaty structure, examines each of the five annexed protocols, and assesses how the regime has performed in the four decades since it entered into force. The study uses doctrinal legal analysis combined with a review of recent scholarship in law, security studies and ethics published between 2021 and 2026. Three findings stand out. First, the strength of the #framework_convention model is flexibility: new protocols can be added without renegotiating the parent treaty, and the ban on #blinding_laser_weapons shows that a #preventive_ban is possible. Second, the weakness of the same model is that flexibility depends on political will, and the #consensus_rule allows a single state to block progress, as the long debate on #lethal_autonomous_weapons_systems demonstrates. Third, the humanitarian centre of gravity has partly moved outside the convention, into the #Ottawa_Convention, the #Convention_on_Cluster_Munitions and the 2022 political declaration on explosive weapons in populated areas. Recent state practice, including the withdrawal of five European states from the mine ban treaty between 2025 and 2026, has made the older and weaker CCW rules on mines newly important again. The article argues that the convention is neither a failure nor a success in simple terms. It is a floor, not a ceiling, and its future value depends on whether states are willing to raise that floor. Keywords: international humanitarian law; conventional weapons; unnecessary suffering; indiscriminate effects; autonomous weapons; humanitarian disarmament; arms control. Introduction Every legal system that regulates violence must answer a hard question. If war is permitted at all, what limits should apply to the way it is fought? The law of #armed_conflict answers this question by accepting that soldiers may be killed, while insisting that the methods used must not go beyond what is needed to defeat the enemy. This is the idea behind the ban on weapons of a nature to cause #unnecessary_suffering. The same body of law also insists that weapons must be capable of being aimed at lawful targets. This is the idea behind the ban on weapons whose effects cannot be limited, which is another way of describing the rule against #indiscriminate_effects. These two ideas are old. They can be traced to the St Petersburg Declaration of 1868, to the Hague Regulations of 1899 and 1907, and to Article 35 of Additional Protocol I to the Geneva Conventions of 1977. They are also broad. A rule that says "do not use weapons that cause unnecessary suffering" is easy to state and hard to apply. Does it ban a rifle bullet that tumbles inside the body? Does it ban a flame weapon? Does it ban a mine that stays active for thirty years? Broad principles do not answer these questions on their own. States therefore need a way to translate general #international_humanitarian_law into concrete rules about concrete objects. The convention adopted in Geneva in 1980 was designed to do exactly that work of translation. Its full name is long, and most scholars, diplomats and students simply call it the #CCW. Some older sources call it the Inhumane Weapons Convention. The treaty itself contains almost no rules about weapons. It is a shell, a framework convention that sets out how states join, how they meet, how new rules are added, and how the whole system connects to the wider body of the law of war. The real content sits in the annexed protocols. A state that joins the convention must accept at least two of them. This design choice, made in 1980 for reasons of political convenience, has shaped everything that has happened since. This article has three purposes. The first is descriptive. Students of international law need a clear account of what the convention says, protocol by protocol, in plain language. The second purpose is analytical. The convention is often praised by diplomats and criticised by humanitarian organisations, and both reactions contain some truth. The article tries to explain why the same treaty can look like a success from one angle and a disappointment from another. The third purpose is forward looking. Between 2022 and 2026 the international environment for #arms_control changed sharply. War returned to Europe on a large scale. Cities were bombarded. Anti personnel mines were laid again by states that had promised never to use them. Weapons with autonomous functions were used in real operations rather than in laboratories. Any honest assessment of the CCW must be written against that background. The argument of the article can be stated simply. The convention works best when it does something narrow and technical, and it works worst when it is asked to do something broad and political. It succeeded in banning weapons that no army badly wanted, such as blinding laser weapons and plastic fragments invisible to X rays. It succeeded in building detailed technical duties for clearing #explosive_remnants_of_war. It has struggled, and continues to struggle, wherever a rule would take away a capability that powerful militaries wish to keep. That pattern is not a moral failing of the treaty. It is a structural feature of a regime that runs on unanimity. Research aims, questions and method 2.1 Aims The study aims to provide an integrated legal and institutional analysis of the CCW that is accessible to students while meeting the standards expected of peer reviewed scholarship. It does not attempt to settle every doctrinal dispute. It attempts to map the regime, to explain the reasoning behind its rules, and to identify the pressure points where law and practice no longer match. 2.2 Research questions The article is organised around five questions. First, what is the legal architecture of the convention, and how do the framework instrument and its protocols fit together? Second, how do the two customary rules on unnecessary suffering and indiscriminate effects operate as legal standards, and what work does the #Martens_Clause do alongside them? Third, what does each of the five protocols actually prohibit or restrict, and where do the definitions leave gaps? Fourth, why has the regime been able to produce agreement in some areas and deadlock in others, in particular on lethal autonomous weapons systems? Fifth, what reforms are realistic, and what should researchers study next? 2.3 Method The method is doctrinal and qualitative. The primary sources are the treaty text and the five protocols, together with official documents of the meetings of #High_Contracting_Parties and of the #Group_of_Governmental_Experts. The secondary sources are peer reviewed books and journal articles, with a strong preference for work published in the last five years, so that the analysis reflects the current state of debate rather than the debate as it stood a decade ago. The article also draws on documented state practice, including national withdrawal decisions and reported uses of particular weapons, in order to test whether the rules on paper match behaviour in the field. The approach has limits that should be stated openly. Doctrinal work cannot measure how many civilian lives a treaty has saved. It cannot see inside classified targeting processes. Where the article makes claims about effects, it relies on published reporting and on the assessments of scholars, and it flags uncertainty rather than hiding it. Historical background: how the convention was born 3.1 The nineteenth century foundations The modern law on weapons begins with a short document. The St Petersburg Declaration of 1868 banned certain small explosive projectiles and stated a principle that has outlived the specific ban: the only legitimate object of war is to weaken the military forces of the enemy, and that object is exceeded by the use of arms which uselessly aggravate the suffering of disabled men. This sentence contains the whole logic of unnecessary suffering. Suffering is not forbidden. Useless suffering is forbidden, because it adds nothing to the military purpose. The Hague Conferences of 1899 and 1907 turned this idea into treaty language and added the ban on projectiles that expand in the human body. They also produced the clause named after the Russian delegate Fyodor Martens. The #Martens_Clause says, in substance, that where treaty rules run out, civilians and combatants remain under the protection of the principles of international law derived from established custom, from the principles of #humanity, and from the dictates of #public_conscience. That clause matters far more than its age would suggest, because every new weapon technology creates a temporary gap in the treaty rules, and the clause tells lawyers what to do in the meantime. 3.2 The gap after 1949 and the pressure of the 1970s The Geneva Conventions of 1949 dealt mainly with the protection of victims, not with the means of warfare. For most of the twentieth century, conventional weapons were regulated almost entirely by unwritten custom. That changed in the 1970s. Two forces came together. The first was medical and humanitarian evidence gathered by the #ICRC about the effects of napalm, of small calibre high velocity bullets, and of fragmentation weapons. The second was the diplomatic conference of 1974 to 1977 that produced the two Additional Protocols to the Geneva Conventions. #Additional_Protocol_I restated the prohibition on weapons of a nature to cause #superfluous_injury and set out the rules of #distinction, #proportionality and #precautions_in_attack that still govern the conduct of hostilities. States were unwilling to negotiate detailed weapon bans inside that conference, so they pushed the question into a separate process. The United Nations convened a conference in Geneva that adopted the convention on 10 October 1980, with three protocols attached. It was the first treaty in more than seventy years to place limits on ordinary battlefield weapons. 3.3 Cold War politics and the shape of the outcome Recent historical scholarship has shown that the negotiation cannot be understood only as a humanitarian project. It was also an episode in Cold War diplomacy, in which states used the language of humanity while managing pressure from allies, from non aligned states and from the emerging network of humanitarian organisations. Work on forum management in conventional arms control during this period shows how states redirected uncomfortable proposals into venues where they could be slowed down or diluted, a pattern that has clear echoes in the modern debate over autonomous weapons (Mantilla, 2023). Two features of the 1980 outcome reflect this politics. The first is that the treaty avoided any verification machinery. The second is that it originally applied only to international armed conflicts, an omission corrected only in 2001, when an amendment to Article 1 extended the whole convention and its protocols to non international armed conflicts for those states that accepted the change. Both features were the price of agreement. The legal architecture of the convention 4.1 A shell and its contents The convention is best pictured as a shell with drawers. The shell contains procedural articles: scope of application, signature, ratification, entry into force, review conferences, amendments, and withdrawal. The drawers are the protocols, each containing rules about one class of weapon. A state that wants to join must accept at least two protocols. It may accept more later. It may also refuse a protocol indefinitely. This creates a patchwork. Two states may both be parties to the convention while owing each other different obligations, because one has accepted #Protocol_III and the other has not. For students, the practical lesson is that the question "is state X bound by the convention?" is never enough. The correct question is always "which protocols has state X accepted, and did it accept the 2001 amendment extending them to internal conflicts?" 4.2 The saving clause Article 1 of the convention makes clear that nothing in the treaty or its protocols detracts from other duties imposed by #international_humanitarian_law. This matters. A weapon that falls outside every protocol is not automatically lawful. It remains subject to the general rules, to the Martens Clause, and to the duty of each state to review new weapons. The convention is a floor, not a ceiling. That single sentence in the treaty carries much of the weight in modern arguments about weapons that the drafters never imagined. 4.3 Review conferences and the ability to grow The convention allows for review conferences and for new protocols to be added. This is its most important structural virtue. When states agreed in 1995 to ban laser weapons designed to blind, they did not have to build a new treaty from nothing. They wrote #Protocol_IV and attached it. When the problem of unexploded ordnance became impossible to ignore, they wrote #Protocol_V in 2003. A #review_conference is therefore not a ritual. It is the mechanism through which the regime can grow. The same mechanism, however, depends on agreement, and agreement inside the convention has traditionally meant #consensus_rule. Nothing in the text compels unanimity for every decision, but the practice of the meetings has hardened into an expectation that outcomes are adopted only when no state objects. That practice explains most of what follows in this article. The two governing principles and the Martens Clause 5.1 Unnecessary suffering and superfluous injury The rule against unnecessary suffering protects combatants. It accepts that soldiers may lawfully be attacked and killed, and it asks a comparative question: does the design of this weapon impose harm that goes beyond what is needed to put a soldier out of action? Applying the rule requires balancing. On one side sits the injury caused, measured by severity, by the difficulty of medical treatment, by the likelihood of permanent disability and by the mortality rate. On the other side sits the military advantage that the weapon offers. A weapon is unlawful when the injury it inflicts is clearly out of proportion to the military purpose it serves. Three points are often missed by students. First, the test concerns the nature and design of the weapon, not the feelings of the victim. A bullet that kills instantly may cause no suffering at all and is perfectly lawful. Second, the test is not about lethality. A weapon designed to maim rather than kill can be unlawful precisely because it inflicts lasting harm without adding military value. Third, the test is objective and general. It asks what the weapon does by its nature in normal use, not whether one soldier used it cruelly on one occasion. 5.2 Weapons that cannot discriminate The second rule protects civilians. It prohibits weapons that cannot be directed at a specific military objective, and weapons whose effects cannot be limited as required by law. This is not the same as the rule against indiscriminate attacks. An attack can be indiscriminate even when the weapon is perfectly precise, because the attacker aimed carelessly. A weapon is inherently indiscriminate when no amount of care by the user can confine its effects. The principles of #distinction and #proportionality apply to attacks; the rule discussed here applies to the object itself. Mines illustrate the difference. A mine can be laid with care, mapped, fenced and cleared. The same mine, laid without records in a field beside a village and left there for twenty years, becomes a machine that cannot tell a soldier from a child. This is why the law on mines has moved steadily from rules about use towards rules about design, self destruction and detectability. 5.3 The role of the Martens Clause The Martens Clause is often described as a safety net. It does not by itself prohibit any particular weapon, and reasonable lawyers disagree about how much independent force it carries. Its practical function is to place a burden on the argument that "what is not forbidden is permitted." In debates on new technologies, the clause allows states and civil society to argue that a weapon offends the principles of humanity and the dictates of #public_conscience even before any specific treaty rule exists. The history of #Protocol_IV shows this clearly: the idea that deliberately blinding a soldier for life is unacceptable was not derived from any pre existing text. It came from a shared moral judgement that was then written into law. The five protocols examined 6.1 Protocol I: non detectable fragments The first protocol is one paragraph long. It prohibits the use of any weapon whose primary effect is to injure by fragments which cannot be detected in the human body by X rays. The mischief is easy to understand. Metal fragments can be found and removed. Fragments made of glass or plastic cannot, and a wounded person may carry them, and the infection and pain they cause, for life. The military value of choosing an undetectable material instead of a detectable one is close to zero. This is the clearest possible case of superfluous injury: extra suffering for no military gain. #Protocol_I is often called a symbolic provision because no state has admitted to fielding such a weapon. That criticism misses the point. The protocol shows how the general principle can be turned into an operational rule with a testable criterion. It also shows the political economy of weapons law. Prohibition is easy when nobody wants the weapon. 6.2 Protocol II and Amended Protocol II: mines, booby traps and other devices The original Protocol II of 1980 regulated the use of mines, #booby_traps and similar devices. It required a party laying mines to record their location in some circumstances, restricted remotely delivered mines, and banned certain booby traps, such as those attached to children's toys, to the dead or to medical supplies. Its rules were weak, applied only to international conflicts, and contained no reliable duty to clear what had been laid. The humanitarian consequences of #landmines in the 1980s and early 1990s made those weaknesses impossible to defend. The result was the #Amended_Protocol_II of 1996, in force since 1998, which is a substantially different instrument. It extends to non international armed conflicts. It requires detectability, so that mines can be found by standard equipment. It requires self destruction and self deactivation features for remotely delivered anti personnel mines, so that a mine does not remain lethal for decades. It prohibits the use of mines against civilians and prohibits indiscriminate use. It places responsibility for clearance and for #mine_clearance related record keeping on the party that laid the mines. It also created an annual meeting of states parties and a national reporting requirement, which is one of the few #compliance mechanisms in the whole convention. The protocol still permits anti personnel mines that meet its technical standards. That is why humanitarian campaigners left the convention in the mid 1990s and built the #Ottawa_Convention outside it. For twenty five years, the amended protocol looked like a historical relic, a weaker rule superseded by a stronger ban. Between 2025 and 2026 that changed. Five European states, Estonia, Latvia, Lithuania, Finland and Poland, completed their withdrawal from the mine ban treaty, citing the threat from Russia and the lessons of the war in Ukraine (Human Rights Watch reporting; House of Commons Library briefing). Those states remain parties to the CCW and to #Amended_Protocol_II. In other words, the weaker rule has become, once again, the operative rule for a group of states in the middle of Europe. This is a striking example of #norm_erosion in one regime restoring the practical significance of another. 6.3 Protocol III: incendiary weapons #Protocol_III regulates #incendiary_weapons, meaning weapons primarily designed to set fire to objects or to cause burn injury to persons through flame or heat produced by a chemical reaction. Napalm is the classic example. The protocol prohibits making the civilian population as such the object of attack with such weapons. It prohibits attacking any military objective located within a concentration of civilians by air delivered incendiary weapons. It permits ground launched incendiary attacks on military objectives inside a concentration of civilians in narrow circumstances, where the objective is clearly separated from civilians and all feasible precautions are taken. Two features of this text have generated persistent criticism. The first is the definition. Because the protocol covers weapons "primarily designed" to burn, it arguably excludes multipurpose munitions such as those containing #white_phosphorus, which are formally used to make smoke, to mark targets or to illuminate, even though the burns they cause are just as severe. The second is the difference in treatment between air delivered and ground launched incendiary weapons, which has no obvious humanitarian justification. Human Rights Watch and a number of states parties have argued for closing both gaps. Other lawyers reply that the exclusion of #white_phosphorus is not an accident but a deliberate choice of the drafters, and that the use of such munitions is in any case governed by the general rules on distinction, proportionality and #precautions_in_attack. Documented use of white phosphorus munitions in and around populated areas in recent conflicts, including in Ukraine, Gaza and southern Lebanon, has kept this dispute alive. What is notable for institutional analysis is not the legal disagreement itself but the response of the regime. Proposals simply to set aside meeting time for a discussion of the adequacy of Protocol III have repeatedly failed to gain consensus. A regime that cannot agree to talk about a problem is unlikely to solve it. 6.4 Protocol IV: blinding laser weapons Protocol IV, adopted in 1995, prohibits the employment of laser weapons specifically designed, as their sole combat function or as one of their combat functions, to cause permanent blindness to unenhanced vision. It also prohibits transfer of such weapons and requires all feasible precautions to avoid incidental permanent blindness from the lawful use of other laser systems, such as rangefinders. This protocol is the strongest evidence that the convention can act before harm occurs. The weapons in question were technically feasible and were reaching the point of deployment. The argument that carried the day was moral rather than technical: a soldier deliberately blinded for life, in a world where blindness is a permanent and severe disability, suffers harm that is out of all proportion to the military purpose of putting him out of action. The #ICRC, medical bodies and a group of states led the campaign, and the ban was adopted by consensus. The protocol is frequently invoked as a precedent for a #preventive_ban on autonomous weapons. The comparison is instructive but must be handled carefully. Blinding lasers were a narrow, well defined category of weapon with a clear physical signature and limited military value. Autonomy is not a weapon at all. It is a property that can be attached to almost any system, from an air defence battery to a loitering munition. Scholars who have examined why some anticipatory restrictions succeed while others fail argue that the key variables are how states assess the military advantage of a technology and the vulnerability it creates for them, and that agreements are hollow when states misjudge the technology (Egel, 2025). Applied to autonomy, that framework predicts difficulty, because the perceived advantage is very high and the vulnerability is felt unevenly. 6.5 Protocol V: explosive remnants of war #Protocol_V, adopted in 2003 and in force since 2006, is different in kind from the others. It does not restrict the use of any weapon. It deals with what is left behind. explosive remnants of war means unexploded ordnance and abandoned ordnance: shells that failed to detonate, submunitions lying in fields, stockpiles left in an abandoned base. The protocol places duties on parties after hostilities end. The party in control of territory must mark, clear and destroy explosive remnants in that territory. The party that used the munitions must provide technical, material and financial assistance, and must record and share information about what was fired and where. It also requires warnings and #risk_education for civilians, and it encourages care of victims. It includes an annex of voluntary best practices on the reliability of munitions. For students this protocol is a useful corrective to the idea that weapons law is only about the moment of attack. Much of the civilian harm from modern warfare occurs years later, when a farmer strikes a buried shell. Protocol V is the CCW's most practical humanitarian achievement, even though it is the least discussed. Institutional design: why the regime moves slowly 7.1 The consensus rule Formally, decisions could be taken by vote. In practice, the meetings of the convention operate on the basis that nothing is adopted over the objection of any participating state. This has three consequences. The first is a bias towards the lowest common denominator. Any text must be acceptable to the most reluctant militarily significant state in the room. The second is that procedural steps become substantive battles. If agreeing to hold a discussion requires consensus, then blocking a discussion becomes an easy way to block a rule. The third is a hidden benefit that critics sometimes overlook. Because the major military powers participate, any rule adopted here binds the states that actually possess the weapons. A prohibition adopted by vote against the wishes of those states would be more ambitious on paper and less effective in the field. This tension between ambition and inclusiveness runs through the whole field of #humanitarian_disarmament. 7.2 Weak compliance machinery The convention has no inspectorate, no fact finding body of its own, and no court. Amended Protocol II and Protocol V require national annual reports, which is a form of transparency, but reporting rates are imperfect and reports are self declared. There is no #verification system comparable to those in the chemical weapons regime. Compliance therefore depends on three softer mechanisms: national implementation through military manuals and criminal law; the political cost of being named by other states, by the ICRC and by #civil_society; and the general enforcement architecture of international law, including individual criminal responsibility for war crimes. These mechanisms are real but slow, and they work poorly against a state that has decided that the military need is overwhelming. 7.3 Weapons reviews as the hidden pillar One duty deserves more attention than it usually receives. Under Article 36 of Additional Protocol I, a state must determine, when studying, developing, acquiring or adopting a new weapon or method of warfare, whether its use would be prohibited by international law. #Article_36_reviews are, in effect, the point at which the abstract principles of unnecessary suffering and indiscriminate effects are applied to actual hardware before it reaches the battlefield. The system has obvious weaknesses. Reviews are national, secret, and conducted by the same government that wants the weapon. Only a minority of states have a formal review mechanism at all. Scholars working on machine learning systems have shown a further problem: if a system continues to learn after deployment, the object reviewed is not the object used, and the traditional idea of a one time #legal_review no longer fits (McFarland and Assaad, 2023). Any serious reform of the CCW should begin with strengthening #weapons_review practice, because that is where law meets engineering. The convention and its rivals: humanitarian disarmament outside the framework 8.1 The Ottawa and Oslo processes The most important fact about the history of the CCW in the 1990s and 2000s is that its two greatest humanitarian achievements happened somewhere else. Frustrated by the weakness of the mine rules, a coalition of states and non governmental organisations negotiated the Ottawa Convention of 1997, which bans anti personnel mines outright. A similar coalition produced the #Convention_on_Cluster_Munitions in 2008 after the convention failed to agree on a strong rule on cluster munitions. These processes shared a method: leave the consensus forum, negotiate among the willing, adopt a categorical ban, and rely on #stigmatisation to influence the states that stay outside. They worked. They also carried a cost. The states with the largest arsenals mostly stayed out, and the resulting norms had to be defended by moral pressure rather than by treaty obligation. 8.2 The 2022 political declaration on explosive weapons in populated areas The most recent example of this pattern is the political declaration on strengthening the protection of civilians from the humanitarian consequences arising from the use of explosive weapons in populated areas, endorsed by eighty three states in Dublin in November 2022 and by further states since. It is not a treaty. It does not create binding rules. It commits endorsing states to restrict or refrain from the use of explosive weapons with wide area effects in populated areas, to collect data, and to assist victims. Analysis of the negotiation shows that the declaration also has implications for the protection of the environment, an area that weapons treaties usually neglect (Bagshaw, 2023). The rise of the #EWIPA_declaration is significant for the CCW in two ways. It confirms that states are willing to act on #protection_of_civilians when they can avoid the veto point of consensus. It also confirms that they prefer soft instruments when hard ones are politically impossible. Whether soft law of this kind builds towards binding rules or provides an excuse to avoid them is one of the central open questions in the field. 8.3 Withdrawal and the fragility of humanitarian norms The events of 2025 and 2026 have complicated the optimistic story of humanitarian disarmament. Lithuania withdrew from the Convention on Cluster Munitions, effective in March 2025. Then Estonia, Latvia, Lithuania, Poland and Finland withdrew from the Ottawa Convention, with the withdrawals taking effect between December 2025 and February 2026. The stated reason in each case was the security situation created by Russia's invasion of Ukraine and the perceived need for defensive barriers along the eastern flank of NATO. Humanitarian organisations and the ICRC criticised these steps as a dangerous reversal. Defenders argued that the states involved face an existential threat and that mines can be used lawfully if properly recorded and cleared. For the purposes of this article the point is analytical rather than partisan. #treaty_withdrawal is lawful when the treaty allows it, and the mine ban treaty does allow it, subject to notice periods. What the episode reveals is that categorical humanitarian bans are strong when security pressure is low and fragile when it is high. The rules that remain in force for these states are precisely the older, weaker, more permissive rules of Amended Protocol II. The convention that campaigners once dismissed as inadequate has become the last line of legal defence. Emerging technologies: the debate on autonomous weapons 9.1 How the issue entered the regime Since 2014 the convention has been the main multilateral venue for discussing weapons that can select and engage targets without human intervention. Informal expert meetings became, in 2016, a formal Group of Governmental Experts. The group agreed in 2019 on eleven guiding principles, of which the most important state that international humanitarian law applies fully to all weapons systems including these, that responsibility for use rests with humans and cannot be transferred to a machine, and that human machine interaction must ensure compliance with the law. These principles are a real achievement in the sense that all the major powers accepted them. They are also, on their own, not much more than a restatement of existing law. The hard questions are the ones the group has not answered: how much human control is enough, whether any category of these systems should be prohibited outright, and what form any instrument should take. 9.2 The state of the argument Broadly, three camps exist. A large group of states, supported by the ICRC and by a wide #civil_society coalition, favours a legally binding instrument built on a #two_tier_approach: prohibit systems that cannot be used with #meaningful_human_control or that target human beings directly, and regulate the rest through requirements of #predictability, #reliability and human supervision. A second group, including several major military powers, argues that existing law is sufficient, that national #weapons_review practice can manage the risks, and that a ban would be premature and unverifiable. A third group is undecided or hedging. Comparative studies of the practices of China, Russia and the United States show that the positions taken in Geneva track the technological and strategic interests of the states involved, and that in the absence of agreed norms the day to day practice of using systems with autonomous features is itself creating standards for what counts as acceptable human control (Bode, Huelss, Nadibaidze, Qiao-Franco and Watts, 2023; Bode and Huelss, 2022). This is the crucial insight of the norms literature. Law is not the only thing that regulates behaviour. Practice quietly regulates it too, and if the law is silent for long enough, practice will fill the space. 9.3 Deadlock and the move to New York The predictable result of combining a consensus rule with a technology that powerful states wish to develop has been slow progress. In December 2024 the United Nations General Assembly adopted a resolution on autonomous weapons with overwhelming support, referring to the possibility of a two tier approach of prohibitions and regulations. The mandate of the expert group inside the convention was renewed for further sessions in 2025. The shift of energy towards the #UN_General_Assembly matters. It repeats the historical pattern seen with mines and cluster munitions: when the consensus forum stalls, states that want rules look for a forum where a majority can act. Whether this leads to a separate treaty on autonomous weapons, negotiated outside the convention and without the major producers, is the single most important open question about the future of the regime. 9.4 Analytical cautions Two cautions are worth stating for student readers, because the literature on this subject can be breathless. First, definitions carry enormous weight and are contested. If "autonomous weapon system" is defined broadly, it captures air defence systems that have existed for decades. If it is defined narrowly, it captures almost nothing that exists today. Much apparent disagreement about policy is actually disagreement about definitions. Second, some scholars warn that the debate is distorted by hype, and that treating these systems as a science fiction problem obscures the mundane and immediate issues of #targeting error, bias in training data, and the erosion of human judgement in decision support tools (Wood, 2024). A risk based approach, which sorts systems by the severity and probability of the harm they may cause rather than by their label, has been proposed as a more workable regulatory model (Blanchard, Novelli, Floridi and Taddeo, 2025). Questions of #accountability under international criminal law remain unresolved, since a machine cannot be a defendant and the humans in the chain may be many steps removed from the harm (Kraska, 2021; Hassan and Osman, 2023). Contemporary stress tests 10.1 Urban warfare Conflicts in Ukraine, Gaza, Syria, Sudan and Yemen have all been fought to a significant degree inside cities. #urban_warfare exposes the limits of a weapons law built around open battlefields. When artillery with a wide margin of error is used in a dense neighbourhood, the weapon is not inherently indiscriminate, but the effect can be devastating and predictable. The convention has little to say about this. The #EWIPA_declaration was created precisely because the treaty regime had no answer. 10.2 Incendiary effects Reported uses of white phosphorus munitions in populated areas have generated sustained argument about whether Protocol III is fit for purpose. The regime's inability even to schedule dedicated discussions on the question is a measure of institutional weakness rather than legal ambiguity. 10.3 The return of the mine The withdrawals from the mine ban treaty described above create a live legal question that students should follow closely. A withdrawing state remains bound by #customary_international_law, by the general rules on distinction and precautions in attack, and by Amended Protocol II if it is a party. That means detectability requirements, restrictions on remotely delivered mines, duties to record and clear, and a prohibition on directing mines against civilians. The gap between the mine ban treaty and the amended protocol is therefore not the difference between prohibition and lawlessness. It is the difference between a categorical ban and a conditional permission. Whether the conditions will be honoured in a large war is exactly what the next years will test. 10.4 Drones and loitering munitions The widespread use of armed drones and loitering munitions has blurred the line between remotely operated and autonomous systems. Existing law applies to them in full: they must be capable of discriminating, their use must satisfy proportionality, and feasible precautions in attack must be taken. The difficulty is practical. Cheap systems proliferate quickly, are used by non state actors, and rarely undergo any legal review at all (Niyitunga, 2023). No protocol of the convention addresses them directly. Critical assessment 11.1 What the convention does well It converts principle into rule. The general standards of unnecessary suffering and indiscriminate effects are, on their own, too abstract to guide a procurement officer. The protocols turn them into testable criteria: is the fragment visible on an X ray; does the mine self destruct; does the laser have blinding as a combat function. It includes the powerful. Unlike the Ottawa and Oslo instruments, the convention has as parties the states that hold the largest and most advanced arsenals. Any norm accepted here is accepted by the states that matter militarily. It can grow. The framework structure allows new protocols. It has produced one preventive ban, one genuinely strong revision, and one detailed post conflict clearance regime. It provides a permanent forum. Even when no rule is adopted, the annual meetings keep the issues on the diplomatic agenda, generate documents, and give civil society a place to make arguments. Institutions have value simply by existing. 11.2 What the convention does badly It is captured by consensus. One objecting state can block a rule, a discussion, or even a paragraph in a final report. This is the single most important explanation of the regime's slowness. Its definitions are drawn narrowly. The "primarily designed" formula in Protocol III is the clearest example, but the same drafting instinct appears elsewhere. Narrow definitions were the price of agreement in 1980; they are the source of the gaps criticised today. It lacks teeth. Without verification or independent fact finding, allegations of violation become political disputes. It has been outflanked. The most important humanitarian rules on conventional weapons in the last thirty years were made outside the convention. This raises the uncomfortable question of what the convention is now for. 11.3 A balanced judgement The fairest assessment is that the convention is a floor. It sets minimum standards that bind almost every significant military power, and it keeps a permanent conversation alive. It is not, and was never designed to be, the place where ambitious prohibitions are born. The mistake made by both its defenders and its critics is to judge it against the wrong standard. Diplomats err when they present procedural continuity as substantive progress. Campaigners err when they treat the floor as worthless simply because it is low. Recent events, in which the higher ceilings built outside the convention have started to crack under security pressure, are a reminder that a floor is worth defending. Reform pathways 12.1 Reforming the decision rule The most direct reform is to loosen the practice of consensus rule, at least for procedural questions such as the allocation of meeting time. If states cannot agree to discuss a subject, no substantive outcome is possible. A modest change here would cost little and unlock a great deal. It is also the reform that objecting states are most likely to resist, precisely because it works. 12.2 Fixing Protocol III Two amendments would address the criticisms of the incendiary weapons rules. The first would replace the "primarily designed" test with an effects based definition, so that any munition producing severe burn injury through flame or heat falls within scope regardless of its intended purpose. The second would apply the same restriction to ground launched and air delivered incendiary weapons in concentrations of civilians. Both changes are technically simple. Both are politically hard. 12.3 Strengthening weapons review States should be encouraged, and ideally required by protocol, to establish a formal #Article_36_reviews mechanism, to publish the methodology if not the conclusions, and to repeat reviews when a system changes materially after deployment. For learning systems, a single review at procurement is not enough. This is one of the few areas where progress is possible without confronting the hardest political questions, because a review duty does not by itself prohibit anything. 12.4 An instrument on autonomy If a binding instrument on autonomous systems is achievable inside the convention, it will most likely take the two tier form: prohibit systems that cannot be used with #meaningful_human_control and systems designed to target people directly; regulate everything else through obligations of predictability, reliability, traceability, human supervision, deactivation capability, and limits on the duration, geographical scope and scale of operation. If it is not achievable inside the convention, the historical pattern suggests that a coalition will build it outside, and the regime will lose its central role in the most important weapons debate of the century. 12.5 Universalisation and implementation More than sixty states remain outside the convention. #universalisation matters less for its own sake than for the credibility of the regime, and the same is true of Protocol V, whose clearance duties are of direct practical value to states emerging from conflict. Support for national implementation, including model legislation and training of armed forces, is unglamorous work that produces measurable humanitarian returns. Implications for students and a future research agenda For students, four lessons generalise beyond this treaty. First, read the architecture before reading the rules. Whether a regime requires consensus, whether it allows selective ratification, and whether it has a review mechanism will often predict its output better than the elegance of its principles. Second, watch the definitions. Most weapons law disputes are definitional. Whoever controls the definition controls the scope of the prohibition. Third, treat law and practice as two systems that interact. Norms are not only written; they are also enacted. Sustained practice by powerful actors will eventually reshape the legal expectations of everyone else (Bode, 2024). Fourth, remember that treaties are reversible. The withdrawals of 2025 and 2026 should end any assumption that humanitarian progress is a one way ratchet. Several research questions deserve attention. How do Article 36 reviews actually function in states that conduct them, and can any transparency be achieved without harming security? Does the EWIPA declaration change targeting behaviour, and how would we measure that? What is the effect of the mine ban withdrawals on civilian casualty rates and on clearance obligations under Amended Protocol II? Does the migration of the autonomy debate to the UN General Assembly strengthen or weaken the prospect of a rule that binds the major producers? Comparative work on European positions suggests that even within one region the answers vary widely (Badell and Schmitt, 2022), and the wider crisis in arms control forms the backdrop against which all of these questions must be studied (Wisotzki and Kuhn, 2022; Kuhn and Williams, 2025). Conclusion The Convention on Certain Conventional Weapons is a modest treaty with an immodest task. It took two abstract principles of humanity, the ban on weapons that cause unnecessary suffering and the ban on weapons with indiscriminate effects, and tried to turn them into workable rules about mines, fire, fragments, lasers and unexploded shells. It did so through a flexible framework that could grow, and it kept the world's strongest militaries inside the room while doing it. Judged by the standard of what humanitarian advocates hoped for, the record is thin. The most ambitious prohibitions of the modern era were achieved outside the convention, and its central debate on autonomy has been stuck for a decade behind a procedural veto. Judged by the standard of what a consensus based institution containing every major military power could realistically deliver, the record is more respectable: a preventive ban that has never been violated, a rebuilt mine protocol with real technical standards, and a clearance regime that continues to save lives long after ceasefires. The events of the last few years have changed the frame of the argument. When states begin to leave the strongest humanitarian treaties, the value of a widely accepted minimum standard rises. The convention is that minimum standard. Its future usefulness depends on two things: whether its members are willing to reform the rule that lets one state stop everyone, and whether the principle that gave the treaty its name, that suffering in war must never exceed what military necessity truly requires, retains its hold on the public conscience. Law can encode that principle. It cannot, by itself, keep it alive. Hashtags #Convention_on_Certain_Conventional_Weapons #CCW_1980 #Inhumane_Weapons_Convention #Geneva_1980 #laws_of_war #IHL_studies #weapons_law #conventional_weapons_regulation #excessively_injurious #humanitarian_law_research #disarmament_studies #protection_of_civilians_in_war #killer_robots_debate #mines_and_ERW #international_law_students References Badell, D., and Schmitt, L. (2022). Contested views? Tracing European positions on lethal autonomous weapon systems. European Security, 31(2), 242-261. Bagshaw, S. (2023). The 2022 Political Declaration on the Use of Explosive Weapons in Populated Areas: A tool for protecting the environment in armed conflict? International Review of the Red Cross, 105(924), 1293-1312. Blanchard, A., Novelli, C., Floridi, L., and Taddeo, M. (2025). A risk-based regulatory approach to autonomous weapon systems. Digital Society, 4(1). Bode, I. (2024). Emergent normativity: Communities of practice, technology, and lethal autonomous weapon systems. Global Studies Quarterly, 4(1). Bode, I., and Huelss, H. (2022). Autonomous Weapons Systems and International Norms. Montreal: McGill-Queen's University Press. https://doi.org/10.1515/9780228009245 Bode, I., Huelss, H., Nadibaidze, A., Qiao-Franco, G., and Watts, T. F. A. (2023). Prospects for the global governance of autonomous weapons: Comparing Chinese, Russian, and US practices. Ethics and Information Technology, 25(1), Article 5. https://doi.org/10.1007/s10676-023-09678-x Egel, N. (2025). Anticipatory governance and new weapons of war: Lessons from the Convention on Certain Conventional Weapons. Journal of Strategic Studies, 48(3), 720-743. https://doi.org/10.1080/01402390.2024.2442683 Hassan, F. M., and Osman, N. D. (2023). AI-based autonomous weapons and individual criminal responsibility under the Rome Statute. Journal of Digital Technologies and Law, 1(2). Kraska, J. (2021). Command accountability for AI weapon systems in the law of armed conflict. International Law Studies, 97, 407-447. Kuhn, U., and Williams, H. (2025). Governing the impact of emerging technologies: Actors, technologies, and regulation. Journal of Strategic Studies, 48(3), 559-570. https://doi.org/10.1080/01402390.2025.2472213 Mantilla, G. (2023). Deflective cooperation: Social pressure and forum management in Cold War conventional arms control. International Organization, 77(3), 564-598. McFarland, T., and Assaad, Z. (2023). Legal reviews of in situ learning in autonomous weapons. Ethics and Information Technology, 25(1). Niyitunga, E. B. (2023). Armed drones and international humanitarian law. Digital Policy Studies, 1(2). Payne, K. (2021). I, Warbot: The Dawn of Artificially Intelligent Conflict. Oxford: Oxford University Press. Rosert, E., and Sauer, F. (2021). How (not) to stop the killer robots: A comparative analysis of humanitarian disarmament campaign strategies. Contemporary Security Policy, 42(1), 4-29. Sassoli, M. (2024). International Humanitarian Law: Rules, Controversies, and Solutions to Problems Arising in Warfare (2nd ed.). Cheltenham: Edward Elgar Publishing. Scharre, P. (2023). Four Battlegrounds: Power in the Age of Artificial Intelligence. New York: W. W. Norton. Wisotzki, S., and Kuhn, U. (2022). Crisis in arms control: An introduction. Zeitschrift fur Friedens- und Konfliktforschung. https://doi.org/10.1007/s42597-022-00074-8 Wood, N. G. (2024). Regulating autonomous and AI-enabled weapon systems: The dangers of hype. AI and Ethics, 4(3), 805-817. Primary legal instruments and official documents Convention on Prohibitions or Restrictions on the Use of Certain Conventional Weapons Which May Be Deemed to Be Excessively Injurious or to Have Indiscriminate Effects, Geneva, 10 October 1980, and its annexed Protocols I to V. Protocol on Non-Detectable Fragments (Protocol I), 1980. Protocol on Prohibitions or Restrictions on the Use of Mines, Booby-Traps and Other Devices (Protocol II), 1980, as amended on 3 May 1996. Protocol on Prohibitions or Restrictions on the Use of Incendiary Weapons (Protocol III), 1980. Protocol on Blinding Laser Weapons (Protocol IV), 1995. Protocol on Explosive Remnants of War (Protocol V), 2003. Protocol Additional to the Geneva Conventions of 12 August 1949, and Relating to the Protection of Victims of International Armed Conflicts (Additional Protocol I), 1977, Articles 35, 36, 48, 51 and 57. Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on Their Destruction, Oslo, 18 September 1997. Convention on Cluster Munitions, Dublin, 30 May 2008. Political Declaration on Strengthening the Protection of Civilians from the Humanitarian Consequences Arising from the Use of Explosive Weapons in Populated Areas, Dublin, 18 November 2022. Group of Governmental Experts on Emerging Technologies in the Area of Lethal Autonomous Weapons Systems, Guiding Principles affirmed by the Group and endorsed by the High Contracting Parties, and subsequent working papers and rolling texts, United Nations, Geneva, 2019 to 2025. United Nations General Assembly, Resolution on lethal autonomous weapons systems, adopted 2 December 2024.

The #Convention_on_Cluster_Munitions (CCM), adopted in Dublin on 30 May 2008 and in force since 1 August 2010, prohibits the use, stockpiling, production and transfer of #cluster_munitions in all circumstances, and adds positive duties on #stockpile_destruction, land #clearance, #risk_education and #victim_assistance. This article offers a structured assessment of the treaty for students of international law and security studies. It traces the weapon's technical character and its record of #civilian_harm, reconstructs the #Oslo_Process that produced the ban, and analyses the treaty's legal architecture article by article. It then measures performance against evidence: States Parties have destroyed effectively all of their declared stocks, clearance operators release large areas of contaminated land each year, and no State Party has been credibly accused of new use since 2008. Against these gains stand serious strains. States outside the treaty continue to use the weapon in Ukraine, Myanmar and elsewhere; the United States transferred cluster munitions to Ukraine in several tranches from 2023; and in March 2025 Lithuania became the first state ever to leave a #humanitarian_disarmament treaty. The article argues that the CCM should be understood as a partial but functioning prohibition regime whose legal core is strong, whose humanitarian obligations are underfunded, and whose stigmatising power is now being tested by a security environment that rewards short-term military argument over long-term civilian protection. It concludes with recommendations on #universalisation, on the interpretation of the assistance prohibition, on withdrawal safeguards, and on the financing of #survivor support ahead of the treaty's review conference in 2026. Keywords: cluster munitions; humanitarian disarmament; international humanitarian law; victim assistance; explosive remnants of war; treaty compliance; norm erosion 1. Introduction Few weapons have travelled as quickly from routine military inventory to prohibited status as the cluster munition. For most of the second half of the twentieth century, armed forces in Europe, North America, the Soviet Union and beyond treated the weapon as an ordinary tool of area fire. It was cheap, it covered ground, and it saturated targets that were spread out or moving. By 2008, more than one hundred states had agreed that this same weapon was unacceptable in any circumstance. That reversal is the subject of this article. The #Convention_on_Cluster_Munitions was adopted at a diplomatic conference in Dublin on 30 May 2008, opened for signature in Oslo on 3 December of the same year, and entered into force on 1 August 2010 once thirty states had ratified it. Its central provision is short and unqualified. States Parties undertake never, under any circumstances, to use, develop, produce, acquire, stockpile, retain or transfer cluster munitions, and never to assist, encourage or induce anyone else to do any of these things. The treaty does not attempt to regulate the weapon's use, to limit it to certain targets, or to set a permitted failure rate. It removes the weapon from the arsenals of its members. That structure places the CCM in a distinct family of instruments now usually described as #humanitarian_disarmament. The family includes the 1997 Mine Ban Treaty, the 2008 convention studied here, the 2013 Arms Trade Treaty and the 2017 Treaty on the Prohibition of Nuclear Weapons. What links them is a method rather than a subject. Each was negotiated outside consensus-bound forums, each was driven by a coalition of medium-sized states working alongside #civil_society and the #ICRC, and each starts from the observed effects of a weapon on people rather than from the security interests of its possessors (Dunworth and Hood, 2021; Cooper, 2022). The case for studying the CCM now is stronger than at any point since its entry into force. Three developments have converged. First, the war in Ukraine has produced the heaviest #cluster_munition casualty figures recorded since monitoring began, with the weapon used by two states that are not parties to the treaty. Second, the United States, also not a party, transferred cluster munitions to Ukraine in a series of consignments beginning in July 2023, and at least some of these reportedly moved through the territory of a State Party. Third, and most striking, Lithuania completed its #withdrawal from the convention on 6 March 2025, becoming the first state in history to leave a humanitarian disarmament treaty of this kind. Each of these facts raises a distinct legal question, and together they raise a political one: whether the prohibition can hold when the security climate turns. This article pursues four objectives. The first is descriptive. It explains, in accessible terms, what a cluster munition is, why the weapon produces the patterns of harm it does, and what the historical record shows about that harm from the Second World War through Vietnam, Lebanon, Iraq, Syria, Nagorno-Karabakh and Ukraine. The second is doctrinal. It sets out the legal architecture of the treaty: the definition in Article 2, the prohibition in Article 1, the timelines in Articles 3 and 4, the rights-based duties in Article 5, the cooperation and reporting duties in Articles 6 and 7, the compliance machinery in Article 8, the national implementation duty in Article 9, and the two provisions that generate most argument, namely Article 21 on relations with states not party and Article 20 on withdrawal. The third is empirical. It assesses what the treaty has actually achieved by drawing on the annual reporting of the #Cluster_Munition_Monitor and on operator reporting from mine action agencies, covering stockpile destruction, land release, casualty trends and #victim_assistance. The fourth is analytical. It asks what current pressures mean for the durability of the prohibition, and whether the concept of #norm_erosion accurately captures what is happening, or whether the picture is better described as a stable legal core surrounded by contested outer layers. The article proceeds as follows. Section 2 reviews the literature. Section 3 sets out methodology and its limits. Section 4 explains the weapon and its humanitarian record. Section 5 reconstructs the negotiation. Section 6 analyses the treaty text. Section 7 presents the implementation record. Section 8 examines victim assistance. Section 9 analyses the contemporary challenges. Section 10 applies norm theory. Section 11 discusses the findings, Section 12 offers recommendations, and Section 13 concludes. 2. Literature review Scholarship on the CCM can be grouped into four streams, each answering a different question. 2.1 The legality stream The first stream asks whether cluster munitions are lawful outside the treaty. This is not a trivial question, because the states now using the weapon are precisely those that never joined. Poposka and Nuredin (2024) provide a careful recent statement of the orthodox position. They argue that since Russia, Ukraine and the United States are not parties, the CCM does not bind them as #treaty_law, and that the general rules of #customary_international_law do not contain a categorical ban on the weapon as such. What customary law does contain, they note, are the prohibitions on weapons that are indiscriminate by nature and on means of warfare that cause superfluous injury, together with the general rules on #principle_of_distinction, #proportionality, #military_necessity and precautions. On their analysis, the lawfulness of any given cluster munition attack must be judged case by case, and many documented attacks in populated areas fail that test decisively. This position is uncomfortable for advocates but analytically honest, and most legal commentary converges on it. The disagreement lies in emphasis. One group stresses that a weapon with wide area effects and a high failure rate can almost never satisfy distinction and proportionality when fired into a town, so that the case-by-case test collapses into a near-ban in practice. Another group, associated with military legal writing, stresses that the same weapon fired at massed armour on open ground may be perfectly lawful, and that expanding proportionality to capture long-term effects from unexploded remnants goes beyond settled state practice. Students should notice that both camps accept the same legal framework and differ on how much weight the post-conflict harm carries inside it. 2.2 The humanitarian disarmament stream The second stream treats the CCM as an example of a method. Dunworth and Hood (2021) situate the convention within a broader field of disarmament law, and Docherty (2021) argues that the defining innovation of these treaties is not the prohibition itself but the package of positive obligations attached to it. On this view, a ban that also requires clearance, stockpile destruction, risk education and assistance to victims is qualitatively different from a ban that simply forbids. It creates an ongoing programme of work, a calendar of meetings, and a set of measurable outputs that give the norm continuous visibility. Cooper (2022) is more sceptical, noting that these regimes tend to exempt the overwhelming majority of military technology from scrutiny and can function as a moral offset, allowing states to demonstrate humanitarian credentials at a low strategic price. 2.3 The victim assistance stream The third stream focuses on the treaty's most original contribution. Docherty and Sanders-Zakre (2023) trace how #victim_assistance developed from a soft aspiration in the Mine Ban Treaty into a legal obligation in the CCM, shaped decisively by the 2006 Convention on the Rights of Persons with Disabilities. They show that the CCM's Article 5 embeds a rights-based approach: it defines victims broadly to include families and affected communities, it requires non-discrimination, and it demands the participation of survivors in the design of the programmes meant to serve them. Their assessment of implementation is mixed, and the gap between the legal standard and the funded reality is a recurring theme in this literature. Clinical research supports that concern from another direction. Weston and colleagues (2025) document how care for explosive ordnance casualties in low-resource settings remains constrained by weak surgical capacity, limited rehabilitation and the absence of context-appropriate guidance. 2.4 The contamination and public health stream The fourth stream is empirical and epidemiological. It examines the burden of #explosive_remnants_of_war on populations long after fighting ends. Torelli (2023) compares patterns of explosive violence harm in Ukraine before and after February 2022 and finds a sharp change in scale and in the exposure of civilians in urban areas. Older but foundational work in this stream established the pattern that later CCM advocacy relied on: that submunition casualties are concentrated among agricultural workers, scrap collectors and children, and that they continue for decades. 2.5 The gap What the literature does not yet contain, and what this article attempts, is an integrated student-facing assessment that connects the legal text, the implementation data and the 2023 to 2026 stress period in one analysis. The Lithuanian withdrawal is too recent to have generated a settled scholarly treatment, and most existing commentary on it is found in policy briefings and legal blogs rather than in peer-reviewed work. 3. Methodology This is a doctrinal and evaluative study. It combines legal analysis of the treaty text with a review of secondary literature and of published implementation data. Three categories of source are used. The first is primary legal material: the text of the #Convention_on_Cluster_Munitions itself, the action plans adopted at its review conferences, and the reports of its meetings of States Parties. The second is scholarly literature, drawn where possible from work published in the last five years. The third is systematic monitoring data, chiefly the annual #Cluster_Munition_Monitor produced by the Cluster Munition Coalition, along with operator reporting from mine action agencies. Three limitations should be stated plainly. First, the monitoring data on which any assessment of #compliance depends is produced substantially by civil society organisations that campaigned for the treaty. This does not make the data unreliable, and the Monitor's methodology is transparent and widely accepted, including by governments. But readers should understand that the treaty's verification regime is not a formal inspectorate; it is a civil society reporting system supplemented by the self-reporting of States Parties under Article 7. Second, casualty figures from active conflicts are almost certainly undercounts. Reporting depends on access, and access is worst where fighting is heaviest. Figures cited in this article should be read as recorded minimums rather than as totals. Third, this article does not attempt to adjudicate the military utility of the weapon. That question requires operational data that is not publicly available, and both proponents and opponents of the ban tend to assert conclusions on it that their evidence does not support. 4. The weapon and its humanitarian record 4.1 What a cluster munition is A #cluster_munition is a container. It is delivered by aircraft, rocket, missile or artillery shell, opens in flight, and releases a payload of smaller explosive #submunitions, sometimes called bomblets. A single container may hold dozens or hundreds of them. A rocket salvo may put thousands into the air at once. The military rationale is straightforward. Against a target that is dispersed, moving, or imprecisely located, such as an infantry formation, an artillery battery or a vehicle column, spreading many small explosions across an area is more efficient than delivering one large explosion at a single point. The same logic applies to denying the use of an airfield: one large crater is quickly repaired, while several hundred small ones are not. Two design characteristics generate the humanitarian problem. The first is the #wide_area_effects. Submunitions do not land on a point. They fall across a footprint that may span the area of several football fields, and they cannot be aimed individually once released. Where the footprint includes a military objective and a residential street, the weapon does not distinguish between them. This is why the weapon is so often described as indiscriminate in effect when used in a populated area, even when it is lawfully aimed. The second is the #dud_rate. Submunitions are meant to explode on impact. A significant proportion do not. Fuzes fail, submunitions land in mud, snow, water or vegetation that cushions the impact, or they are released outside their design envelope. Historical failure rates in operational use have often been far higher than the rates claimed in manufacturers' testing. Each failed submunition becomes, in effect, an antipersonnel mine that nobody mapped. It sits in a rice field, an olive grove, a garden or a schoolyard, and it remains lethal for years or decades. The consequence is a weapon that harms civilians twice: during the attack, through area effects, and after the attack, through #unexploded_ordnance. The second harm is what makes cluster munitions different from other artillery, and it is the harm that drove the treaty. 4.2 The pattern of harm The record is long. Cluster munitions were used by Soviet forces against German troops in 1943. They were used massively by the United States in Southeast Asia during the 1960s and 1970s. They were used in Western Sahara, Afghanistan, the Falklands, the 1991 Gulf War, the former Yugoslavia, Chechnya, the Horn of Africa, Kosovo, Iraq, Lebanon, Georgia, Libya, Syria, Yemen, Nagorno-Karabakh and Ukraine. Two cases shaped the political argument. #Laos remains the most heavily contaminated country on earth. During the bombing of the 1960s and 1970s, an enormous tonnage of ordnance was dropped, a substantial share of which failed to function. The consequences did not end with the war. Tens of thousands of people have been killed or injured by explosive remnants since 1964, and roughly half of those casualties occurred after the fighting stopped. Contamination affects thousands of villages and continues to restrict farming, road building, school construction and irrigation. Clearance continues today, more than fifty years after the last aircraft left, and national authorities still destroy tens of thousands of items each year. #Lebanon in 2006 was the immediate trigger. During a conflict lasting about a month, several million submunitions were fired into southern Lebanon, a very large number of which failed to explode. Because the heaviest use came in the final days before the ceasefire, the contamination fell across the exact ground that displaced families were about to return to. Casualties began within days of the fighting ending. Clearance operators are still working in the same villages nearly twenty years later. For diplomats who had been arguing that the existing rules of #international_humanitarian_law were adequate, the Lebanon experience was decisive. The rules had been in force. The harm had happened anyway. 4.3 Who is harmed The demographic pattern is consistent across contexts. Recorded casualties from cluster munitions and their remnants are overwhelmingly civilian. In 2022, monitoring recorded 1,172 new casualties worldwide, the highest annual figure since systematic reporting began, and around ninety-five per cent of those recorded casualties were civilians. In 2023, the civilian share of recorded casualties remained above ninety per cent. Children make up a large and consistent proportion of remnant casualties, because unexploded submunitions are small, sometimes brightly coloured, and often shaped like objects a child will pick up. The injuries are severe. Blast and fragmentation at close range produce traumatic amputation, penetrating abdominal and chest wounds, blindness, deafness and burns. Survivors need emergency surgery that is frequently unavailable nearby, then rehabilitation and prosthetics that most affected countries cannot fund at scale, then psychological support that is almost never available at all (Weston et al., 2025). The economic consequence follows the medical one: a farm worker who loses a leg loses a livelihood, and a family that loses its main earner may lose a generation of schooling. 5. The Oslo Process and the making of the treaty 5.1 Why the existing forum failed Efforts to regulate cluster munitions were pursued for years inside the Convention on Certain Conventional Weapons (CCW). That forum operates by consensus and includes the major producers and users. Its 2003 Protocol V dealt with explosive remnants of war, but it did so after the fact: it created duties to clear and to record, without restricting the weapons that produced the remnants in the first place. Attempts to move further were blocked. By 2006, it was clear to a group of states that consensus procedure gave a veto to precisely the states least willing to act. 5.2 The Oslo alternative In February 2007, Norway convened a conference in Oslo. The forty-six states that attended committed themselves to concluding, by 2008, a binding instrument prohibiting cluster munitions that cause unacceptable harm to civilians. The #Oslo_Process that followed moved through Lima, Vienna and Wellington to the #Dublin_Diplomatic_Conference of May 2008. Three features of the process deserve attention, because they explain the shape of the final text. The process was open to like-minded states and closed to obstruction. It proceeded without the major producers. That was a cost, since it meant the treaty would not bind them, and a benefit, since it meant the treaty would not be diluted by them. The process gave a formal role to #civil_society. The #Cluster_Munition_Coalition, formed in 2003, worked alongside sympathetic governments, the #ICRC and United Nations agencies. Survivors of cluster munition attacks participated directly in the negotiation. This was not decoration. It shaped the definition of a victim and the content of the assistance obligation, and it is one reason the treaty reads as it does. The process was driven by evidence of effects. Delegations were presented with clearance data, casualty data and field testimony rather than with abstract arguments about weapon categories. The result was a treaty organised around observed harm. 5.3 The central bargain of the definition The hardest negotiation concerned the definition. A ban that captured every munition with more than one warhead would have swept in weapons that had never caused the humanitarian pattern the treaty targeted. A ban that allowed states to keep any weapon they described as improved would have been worthless. The solution in Article 2 works by a general rule and a set of cumulative exclusions. The general rule captures conventional munitions designed to disperse or release explosive submunitions each weighing less than twenty kilograms. The exclusions then remove munitions that do not produce the two problem characteristics. Excluded, among others, are munitions that dispense flares, smoke, chaff or electronic effects rather than explosives; munitions designed to produce electrical or electronic effects; and munitions that contain fewer than ten submunitions, where each submunition weighs more than four kilograms, is designed to detect and engage a single target, and is equipped with both an electronic self-destruction mechanism and an electronic self-deactivating feature. Read together, the criteria are demanding, and they are cumulative. A weapon must satisfy all of them to escape the ban. In practice, the definition prohibits the weapons that produce wide area effects and large numbers of duds, and permits a narrow class of sensor-fused munitions that engage individual targets and cannot leave behind live remnants in significant numbers. Students should note the logic here: the treaty does not ban a name, it bans a set of effects. 6. The legal architecture of the Convention 6.1 Article 1: the prohibition Article 1 is the core. Each State Party undertakes never under any circumstances to use cluster munitions; to develop, produce, otherwise acquire, stockpile, retain or transfer them; or to assist, encourage or induce anyone to engage in any activity prohibited to a State Party. Three features matter. The words never under any circumstances exclude the balancing test that governs most of #international_humanitarian_law. There is no military necessity exception, no proportionality assessment, no defence of overwhelming need. For States Parties, the question of lawful use does not arise. The prohibition covers the full life cycle. A state cannot lawfully research, manufacture, buy, hold or move the weapon, not merely fire it. The assistance prohibition in Article 1(1)(c) extends the treaty outward. It is the provision that has generated the most difficult interpretive problems in the treaty's history, because it turns the actions of a State Party's allies into a matter of that State Party's own legal responsibility. Article 1 also applies the prohibition to explosive bomblets specifically designed to be dispersed from an attached dispenser, closing an obvious gap. 6.2 Article 3: stockpiles Article 3 requires each State Party to separate its stockpiled cluster munitions from operational stocks immediately, to mark them for destruction, and to destroy them as soon as possible and no later than eight years after the treaty enters into force for that state. Extensions may be requested and are granted for defined periods with justification. The article permits the retention of a minimum number of cluster munitions and submunitions for the development of, and training in, detection, clearance and destruction techniques, or for the development of counter-measures. The word minimum is doing real work here, and the practice of a small number of States Parties that retain stocks without consuming them has been persistently criticised in monitoring reports and in meetings of States Parties. 6.3 Article 4: clearance Article 4 requires each State Party to clear and destroy cluster munition remnants in contaminated areas under its jurisdiction or control, as soon as possible and no later than ten years after entry into force for it. It requires survey, marking, fencing and #risk_education for civilians pending clearance. Extension requests are permitted, and in practice they are common, because contamination in countries such as Laos is measured in decades of work rather than years. Article 4 also addresses a politically sensitive question: the responsibility of a state that used cluster munitions on the territory of another. The provision encourages, but does not require, the user state to provide assistance for clearance. The distinction between encouragement and obligation was one of the compromises that made agreement possible, and it remains a live grievance for affected states. 6.4 Article 5: victim assistance Article 5 requires each State Party to adequately provide, in accordance with applicable international humanitarian and human rights law, age- and gender-sensitive assistance to #cluster_munition_victims in areas under its jurisdiction or control. That assistance includes medical care, rehabilitation, psychological support, and provision for social and economic inclusion. It requires data collection, national plans and budgets, and, critically, the close consultation and active involvement of victims and their representative organisations. The definition of victim in Article 2 is deliberately wide. It reaches persons killed or injured, but also those who suffered economic loss, social marginalisation or substantial impairment of the realisation of their rights, and it includes affected families and communities. This is the provision in which the influence of the Convention on the Rights of Persons with Disabilities is most visible (Docherty and Sanders-Zakre, 2023). The obligation is not charitable. It is framed as the realisation of rights, and it is owed to people as rights-holders rather than as beneficiaries. Article 5 is also the article most vulnerable to the gap between law and money. It is the only core obligation with no deadline, which reflects the honest recognition that survivor needs are lifelong. It is also, unsurprisingly, the least well funded. 6.5 Articles 6 and 7: cooperation and transparency Article 6 establishes the framework for #international_cooperation and assistance. States Parties in a position to do so are required to provide technical, material and financial assistance to affected States Parties. It also directs assistance towards clearance, stockpile destruction and victim support, and encourages the sharing of technology and information. Article 7 requires annual #transparency_reporting. States Parties must report on stockpiles, destruction, contamination, clearance, national implementation measures, victim assistance programmes, retained munitions and the technical characteristics of the weapons they hold. Reporting forms were adopted at the first meeting of States Parties and revised subsequently. Reporting compliance is imperfect but generally strong, and the reports are the backbone of the treaty's verification system. 6.6 Articles 8 and 9: compliance and national implementation Article 8 creates a facilitation procedure. Where a State Party wishes to clarify a compliance question, it may submit a request for clarification through the Secretary-General, and a meeting of States Parties may be convened to consider the matter. There is no court, no inspectorate and no sanction. Compliance depends on transparency, peer pressure and reputation. Article 9 requires each State Party to take all appropriate legal, administrative and other measures to implement the convention domestically, including the imposition of penal sanctions to prevent and suppress prohibited activity by persons or on territory under its jurisdiction. This is the provision that turns an international obligation into national criminal law, and its uneven implementation is a recognised weakness. A prohibition that exists in a ratification instrument but not in a penal code is harder to enforce against, for example, a domestic manufacturer or a financial institution investing in one. 6.7 Article 21: relations with states not party Article 21 is the treaty's most argued-over provision, and students should read it with care. Paragraph 1 requires each State Party to encourage states not party to join. Paragraph 2 requires each State Party to notify governments of states not party of its obligations, to promote the treaty's norms, and to make its best efforts to discourage states not party from using cluster munitions. Paragraph 3 then provides that States Parties, their military personnel and nationals may engage in military cooperation and operations with states not party that might engage in activities prohibited to a State Party. This is the #interoperability clause. It was inserted at the insistence of states with alliance commitments, above all within NATO, who argued that without it they could not participate in coalition operations alongside forces that retain the weapon. Paragraph 4 sets the limits. Nothing in paragraph 3 authorises a State Party itself to develop, produce, acquire, stockpile, transfer or use cluster munitions, or to expressly request their use where the choice of munitions is under its exclusive control. The tension is obvious. Article 1(1)(c) prohibits assistance with prohibited acts. Article 21(3) permits joint operations with states that commit those acts. Where is the line? The dominant reading among States Parties and commentators is as follows. A State Party may take part in a coalition in which another state uses cluster munitions. It may not itself use, transport, store, finance or specifically request them. Mere participation in a joint headquarters does not breach Article 1, but active facilitation of a specific use does. On this reading, permitting cluster munitions to be transported across one's territory, or storing them on one's soil, or refuelling an aircraft in the knowledge that it will drop them, sits on the wrong side of the line. The practical importance of this reading became clear from 2023 onward, when reporting indicated that United States cluster munitions being transferred to Ukraine had transited through Germany, a State Party. Germany has maintained its own legal position on the matter, and no formal Article 8 clarification proceeding has been brought. The episode nonetheless illustrates that #Article_21 is not a stable settlement but an unresolved argument that becomes urgent whenever an ally goes to war. 6.8 Article 20: withdrawal Article 20 permits a State Party to withdraw. It must give notice to all other States Parties, to the depositary and to the United Nations Security Council, and it must explain its reasons. The withdrawal takes effect six months after receipt of the notification. There is one significant safeguard: if, at the end of that six-month period, the withdrawing state is engaged in an armed conflict, the withdrawal does not take effect before the end of that conflict. The drafters clearly intended this clause to prevent a state from leaving the treaty in order to use the weapon in a war it was already fighting. What they did not anticipate was withdrawal in anticipation of a war that has not started. That is precisely what happened in the Lithuanian case, discussed in Section 9. 7. The implementation record Assessment of any treaty must begin with what it has actually done. On the four core obligations, the record is substantially positive. 7.1 Prohibition compliance Since adoption in 2008, there have been no confirmed reports or credible allegations of new use, production or transfer of cluster munitions by any State Party. That is a significant fact and it deserves to be stated plainly, because it is easy to lose in the noise of contemporary events. Several past users and producers, including France, the Netherlands, South Africa and the United Kingdom, joined the treaty and have adhered to it. The states now using the weapon are, without exception, states that never joined. 7.2 Stockpile destruction This is the treaty's clearest success. States Parties have collectively destroyed close to 1.5 million cluster munitions containing roughly 179 million #submunitions. That figure represents effectively the entirety of the stocks declared by States Parties. Weapons that were destroyed cannot be used, cannot be sold, cannot leak into the arms market and cannot fail to explode in someone's field twenty years from now. The preventive value of this is difficult to overstate and easy to overlook precisely because the harm it prevents never occurs. A residual issue remains. A small number of States Parties retain munitions under the Article 3 exception for training and research, and several of them report consuming none of the retained stock over long periods. Retention without consumption suggests the retained stock is not, in fact, being used for the permitted purpose. Monitoring bodies have repeatedly urged these states to destroy their retained holdings before the review conference in 2026. 7.3 Clearance and land release Clearance is slow, expensive and continuous. It is also working. Contaminated states cleared more than ninety square kilometres in 2022 and more than eighty square kilometres in 2023, destroying tens of thousands of remnants each year. In 2024, States Parties released over one hundred square kilometres of land, the highest annual figure in five years, and destroyed more remnants than in any of the preceding five years. Individual completions matter. Bosnia and Herzegovina finished clearing its cluster munition remnants in August 2023, becoming the ninth State Party to complete its Article 4 obligation. Each completion is a permanent gain. Land that has been surveyed, cleared and released is land that families can farm, build on and cross without fear. The scale of the remaining task should not be minimised. In #Laos, clearance in 2024 covered around seventy square kilometres of agricultural land alone, and national authorities also destroyed thousands of other items of #unexploded_ordnance. Realistic estimates suggest decades of work remain. The Article 4 deadline of ten years is, for the most contaminated countries, an aspiration managed through repeated extension requests rather than a schedule. 7.4 Casualty trends Annual casualties from cluster munitions fell substantially after the treaty's adoption. That decline was reversed by the war in Ukraine. Recorded casualties reached 1,172 in 2022, the highest total since systematic reporting began in 2010, driven overwhelmingly by use in Ukraine. Casualties have continued at elevated levels since, and they arise from two distinct sources: attacks in progress, and remnants left by earlier attacks. The second category will continue to produce casualties in Ukraine for decades regardless of when the fighting stops. The distribution of harm across countries is instructive. Casualties from remnants in recent reporting years have been recorded in Azerbaijan, Iraq, Laos, Lebanon, Mauritania, Yemen and elsewhere, in many cases from contamination laid down decades ago. This is the signature of the weapon: the war ends, the weapon does not. 8. Victim assistance: the treaty's unfinished obligation If stockpile destruction is the treaty's success story, #victim_assistance is its incomplete one. The legal standard, as described above, is genuinely advanced. It is rights-based, it is non-discriminatory, it requires survivor participation, and it treats support as an entitlement rather than a favour (Docherty and Sanders-Zakre, 2023). It has influenced subsequent instruments, including the victim assistance provisions of the Treaty on the Prohibition of Nuclear Weapons. The implementation record is weaker, for reasons that are structural rather than accidental. The obligation falls on the affected state, not on the state that used the weapon. This is legally coherent, since the treaty binds its parties, but its practical effect is that the poorest countries carry the heaviest duty. A state emerging from conflict, with a damaged health system and a collapsed tax base, is asked to fund lifelong rehabilitation for thousands of people. Article 6 obliges states in a position to do so to provide assistance, but the flow of resources is voluntary and it fluctuates with donor politics. The needs are lifelong while the funding is annual. A survivor injured at fifteen may need prosthetic replacement every few years for six decades. Donor cycles do not work that way. Rehabilitation is the strongest area of provision, and mental health support is the weakest. Peer support, which survivors consistently identify as among the most valuable interventions, remains among the least funded. Economic inclusion programmes exist in many affected states but reach a small fraction of those entitled to them. The macroeconomic environment compounds the problem. Health systems in several affected States Parties have been under severe strain, and where the national health system is failing, a specialised programme for one category of injury cannot be insulated from that failure. Clinical evidence points to the same conclusion from a different angle. Weston and colleagues (2025) note that care for explosive ordnance casualties in low-resource settings suffers from an absence of context-appropriate clinical guidance, alongside shortages of surgical and rehabilitative capacity. The problem is therefore not only financial. It is also a problem of systems, standards and trained people. The honest assessment is that Article 5 states an excellent standard that the international community has not resourced. Students evaluating the treaty should hold both facts at once: the norm is strong, and the delivery is thin. 9. The convention under pressure, 2022 to 2026 The last four years have subjected the treaty to the sharpest test of its existence. Four pressures can be distinguished. 9.1 Use by states outside the treaty Russian forces have used cluster munitions extensively in #Ukraine since February 2022, including in populated areas, using both older stocks and newer models. Ukrainian forces have also used them, including in and around Izium in 2022 while the city was under Russian control. Neither state is a party to the treaty. The consequence in law is the case-by-case analysis described in Section 2. The consequence in fact is a mass casualty event that has run for years. The consequence in politics is a reframing of the weapon. From 2008 to 2022, cluster munitions were discussed almost entirely as a humanitarian problem. Since 2022, they have been discussed by many governments as a capability question. That shift in the terms of debate is more dangerous to the treaty than any individual attack. Use has also been documented in #Myanmar, and by former Syrian government forces until the fall of the Assad administration in December 2024. In July 2025, during a border conflict, Thai forces appeared to acknowledge use. None of these states is a party to the convention. The pattern is consistent: the prohibition holds among its members and does not reach beyond them. 9.2 Transfers by a state not party Beginning in July 2023, the United States, which is not a party to the treaty, transferred cluster munitions to Ukraine. Transfers continued in a series of announced consignments through 2024, including 155mm artillery projectiles containing dual-purpose improved conventional munitions and ballistic missiles carrying large numbers of antipersonnel submunitions. Reporting indicated at least seven separate shipments between July 2023 and October 2024. Legally, the United States is bound by neither Article 1 nor Article 21 of the CCM. But the transfers touch the treaty in two ways. First, they cut directly against the norm the treaty was designed to build. A prohibition on #arms_transfers is only as strong as the practice around it, and a major power supplying the weapon to an active conflict tells other states that the transfer of cluster munitions is a normal instrument of security assistance. Second, they raise a genuine legal question for States Parties. If a State Party knowingly permits the transit of cluster munitions across its territory, does that constitute assistance within the meaning of Article 1(1)(c)? Reporting that United States cluster munitions transited Germany, a State Party, sharpened this question considerably. Germany has defended its conduct, and the matter has not been brought to the treaty's formal clarification machinery. The absence of a clarification proceeding is itself notable: the treaty has a compliance mechanism, and its members have chosen not to use it in the hardest case yet presented. That reticence is understandable in the middle of a European war. It is not costless for the norm. 9.3 Withdrawal On 6 March 2025, Lithuania's withdrawal from the convention took effect. It had notified the United Nations in September 2024, following a parliamentary vote in July of that year. Lithuania is the first state ever to leave the CCM, and, setting aside the distinct field of nuclear non-proliferation, the first state ever to leave a humanitarian disarmament treaty of this kind. Lithuania's stated reasons were security-based: deterrence of a possible Russian attack, the need to build defensive capability quickly, interoperability with allies that retain the weapon, the failure of neighbouring states to join the treaty, and claimed technical improvements in munition reliability. The country subsequently announced its withdrawal from the Mine Ban Treaty as well, and several other states on NATO's eastern flank have begun similar processes with respect to landmines. The reaction from States Parties was strongly negative. At the twelfth meeting of States Parties in 2024, around two-thirds of participating states expressed concern, and more than half asked Lithuania to reverse its decision before it took effect. The meeting's final report deeply regretted the decision, urged reconsideration, and condemned the use of cluster munitions by any actor under any circumstances. It reaffirmed that these weapons have proved unreliable, that their effects are indiscriminate, and that any perceived military utility is outweighed by the humanitarian harm they cause. The #ICRC stated that withdrawal from a treaty of international humanitarian law erodes protections in war and damages the fabric of the law itself. Three features of the Lithuanian case deserve analytical attention. The withdrawal used the treaty's own procedure. Lithuania did not breach the CCM; it left it lawfully. This is important. The convention's legal machinery functioned exactly as written. What failed was not the law but the political consensus underneath it. The Article 20 safeguard did not bite. The clause suspending withdrawal during an armed conflict is designed to stop a state from exiting in order to use the weapon in a war it is already in. Lithuania is not at war. It withdrew in anticipation. The safeguard therefore had no application, which suggests a drafting gap rather than a drafting error. The precedent is the danger, not the state. Lithuania holds no cluster munitions and, as of writing, has not been reported to have acquired any. The material effect of its departure on civilian safety today is close to zero. The effect on the treaty's claim to be a settled and irreversible norm is considerable. A prohibition that states may exit when threatened is a different kind of prohibition from one they may not. 9.4 The re-legitimation problem Underlying all three pressures is a fourth, which is discursive. For fifteen years, the CCM succeeded in making cluster munitions unrespectable. Producers lost investors. Militaries removed the weapon from doctrine. States that had never joined nonetheless avoided admitting use. That stigma is now being actively contested. Arguments that were unsayable in 2010 are being made openly in 2025: that the ban was a product of a uniquely peaceful era, that European states disarmed themselves of useful capabilities that their adversaries retained, and that new munitions with lower failure rates deserve a different assessment. These arguments are not absurd, and they should be engaged rather than dismissed. They should also be answered. The claim about improved reliability, in particular, deserves scrutiny. Manufacturers' test-range failure rates have historically been poor predictors of operational failure rates, because operational use involves soft ground, vegetation, extreme temperatures, damaged fuzes, ageing stock and firing outside the design envelope. A munition advertised at one per cent failure has repeatedly performed far worse in war. The treaty's definition already permits genuinely discriminating sensor-fused munitions. The weapons under discussion in the current debate are, for the most part, not those. 10. Analysing the pressure: norm erosion or norm testing? Norm theory offers a useful lens here. The standard account describes a life cycle: a norm emerges, cascades as more states adopt it, and is eventually internalised so that compliance becomes automatic and non-compliance requires justification. The CCM plainly reached the cascade stage. The question is whether it reached internalisation, and whether it is now reversing. The evidence supports a more precise diagnosis than simple #norm_erosion. Consider what has not happened. No State Party has used cluster munitions. No State Party has produced or transferred them. No State Party has been found in breach of Article 1. The treaty gained a new member as recently as 2025. Stockpile destruction is complete. Clearance is proceeding and, by volume, accelerating. On every measure internal to the treaty, the regime is functioning. Now consider what has happened. One member left, lawfully. Non-members are using the weapon at scale. A major power is transferring it. The discourse of legitimate capability is being rebuilt. And States Parties, faced with a hard case involving an allied non-member, have declined to activate their own compliance machinery. The pattern that emerges is not a collapsing norm but a norm whose boundary is being renegotiated. The prohibition itself remains robust for those inside it. What is weakening is the treaty's outward reach: its claim to speak for a general standard of civilised conduct binding on everyone, its power to shame non-members, and its confidence that membership is permanent. This distinction matters for prediction. If the core were failing, we would expect to see States Parties quietly reacquiring the weapon, or refusing to fund clearance, or abandoning reporting. We do not. If the boundary is failing, we would expect to see exactly what we do see: exits at the margin, loud arguments about interoperability, reluctance to confront allies, and a resurgence of military-utility framing. The practical implication is that the treaty's defenders are fighting the wrong battle if they concentrate solely on compliance by members. The members are complying. The threat is that membership stops looking like the obvious choice for a state under pressure. 11. Discussion Four themes emerge from this assessment. 11.1 Prohibition regimes work on their members and only on their members This is banal, but it is constantly forgotten in public debate. The CCM has been extraordinarily effective at eliminating cluster munitions from the arsenals of the states that joined it. It has had no direct legal effect whatsoever on the states that did not. Both halves of that sentence are true, and criticism of the treaty for failing to stop Russian use in Ukraine misunderstands what a treaty is. The relevant question is not whether the treaty stopped non-members but whether the world is better with 111 or 112 states permanently out of the cluster munition business than it would be without them. The clearance data, the destruction data and the casualty data all say yes. 11.2 The positive obligations are the treaty's real innovation and its weakest link Docherty's argument that positive obligations distinguish humanitarian disarmament from classical arms control is persuasive (Docherty, 2021). The CCM does not merely forbid; it repairs. Clearance, risk education and victim assistance turn a prohibition into a programme. But these obligations require money, and money is the thing the treaty cannot compel. Article 6 obliges states in a position to help to help, without specifying how much or to whom. The result is a system where the legal standard for #survivors is high and the funded reality is low. Any serious reform agenda has to begin here, because a repair programme that is not resourced eventually becomes a rhetorical commitment rather than a legal one. 11.3 Interoperability is an unresolved contradiction, not a solved problem Article 21 was a necessary compromise to secure the participation of NATO members. It is also an unstable one. The provision permits joint operations while Article 1 forbids assistance, and the boundary between the two has never been authoritatively settled. In peacetime this ambiguity is tolerable. In wartime, when a State Party's principal ally is supplying the prohibited weapon to an active conflict, the ambiguity becomes a legal and political crisis that States Parties have preferred to manage by silence. Silence is a choice with consequences. Every unchallenged transit, every unexamined storage arrangement, narrows the assistance prohibition by practice rather than by amendment. The treaty's members should decide, deliberately and collectively, what Article 1(1)(c) means before the answer is established for them by events. 11.4 Security pressure beats humanitarian argument in the short term The Lithuanian withdrawal, and the parallel movement out of the Mine Ban Treaty by several states on NATO's eastern flank, show that when a state believes its survival is at stake, a humanitarian prohibition adopted in calmer times is vulnerable. The ICRC's response to this is worth taking seriously: humanitarian law was not written for hopeful days. Its whole purpose is to bind states in exactly the circumstances where they would rather not be bound. The counter-argument that states in Lithuania's position offer is that a rule which no adversary respects is a self-imposed handicap. The honest reply is not to deny the tension but to answer it on its own terms. The military value of cluster munitions is contested even among military professionals. The humanitarian cost is not contested at all. And the state that reintroduces the weapon into its own territory in a defensive war is contaminating the ground its own citizens will have to farm afterwards. 12. Recommendations The following are directed at States Parties, at civil society and at researchers, in advance of the review conference scheduled for 2026. Clarify the assistance prohibition. States Parties should adopt an agreed understanding of Article 1(1)(c) that expressly addresses transit, storage, financing and logistical support. Ambiguity is currently being resolved in favour of permissiveness by default. A collective interpretive statement would resolve it deliberately instead. Use the compliance machinery. Article 8 exists. A clarification request, handled cooperatively and without hostility, would demonstrate that the treaty's members take their own procedures seriously. Continued non-use of the mechanism in the hardest case yet presented sends the opposite signal. Close the withdrawal gap. States Parties cannot amend Article 20 easily, but they can adopt a political declaration recording their understanding that withdrawal in anticipation of conflict is inconsistent with the object and purpose of the treaty, and that a withdrawing state remains bound by clearance and #victim_assistance obligations arising from contamination on its territory. Finance the positive obligations properly. Multi-year, pooled funding for victim assistance and clearance would match the funding cycle to the need cycle. Survivor needs do not reset annually. Resolve the retention question. States Parties retaining cluster munitions without consuming them should destroy them. The Article 3 exception was written for research and training, not for indefinite holding, and a residual stockpile in a State Party is a gift to critics of the treaty. Attack the reliability myth with evidence. States Parties, working with technical institutions, should publish independent assessments of operational failure rates of modern submunitions under realistic field conditions. The improved-reliability argument is currently doing significant political work on the basis of manufacturer claims. Pursue #universalisation where it is achievable. The states most likely to accede are not the major producers. They are affected states, small states and states with no cluster munition capability to give up. Every accession raises the political cost of use, and Vanuatu's accession in 2025 shows the process is not closed. Invest in research. The scholarly literature on the CCM is thin relative to its importance, and the events of 2023 to 2026 have not yet been systematically studied. Priority questions include the operational effect of the weapon in Ukraine, the true failure rates of transferred munitions, the domestic political process that produced the Lithuanian withdrawal, and the long-term outcomes of victim assistance programmes. 13. Limitations This study relies on published monitoring data whose production is dominated by organisations that advocated for the treaty. It uses casualty figures that are recorded minimums drawn from conflict zones with restricted access. It does not evaluate military effectiveness, because the necessary data is not public. And it addresses events, particularly the Lithuanian withdrawal and its consequences, that are still unfolding and have not yet been examined in peer-reviewed scholarship. Conclusions about the trajectory of the norm should therefore be treated as provisional. 14. Conclusion The #Convention_on_Cluster_Munitions did what it set out to do for the states that joined it. It removed the weapon from their arsenals completely. It destroyed nearly 1.5 million munitions and roughly 179 million submunitions. It funded and organised the clearance of land that is now being farmed and built on. It created the strongest #victim_assistance obligation in any weapons treaty, and it wrote survivors into the process that produced it. Since 2008, not one of its members has been credibly accused of using, producing or transferring the weapon. Measured against its own terms, the treaty has succeeded. It did not, and could not, prevent use by the states that stayed outside. The heaviest cluster munition casualties on record were produced after the treaty entered into force, by non-members, in a European war. A major power outside the treaty has transferred the weapon to that war. And in March 2025 a member walked out, lawfully, invoking security. The correct conclusion is neither triumph nor despair. It is that the CCM is a strong legal instrument sitting inside a weakening political consensus. The prohibition's core holds. Its outer boundary, the claim that this weapon is unacceptable to anyone anywhere, is being contested for the first time in fifteen years. Whether that boundary holds depends less on legal drafting than on argument. The states now reconsidering their commitments are not arguing that unexploded #submunitions are harmless. They are arguing that the harm is worth accepting. That is a claim about values under pressure, and it can only be met by insisting on what the evidence has shown for sixty years: that the weapon's costs fall overwhelmingly on people who were not fighting, that those costs continue long after the fighting stops, and that no state which uses cluster munitions on its own territory can protect its own population from what it leaves behind. The review conference in 2026 will be the most consequential meeting in the treaty's history. Its members will decide, in practice if not in words, whether the prohibition on cluster munitions is a permanent achievement of #international_humanitarian_law or a product of a period that has ended. References Bagshaw, S. (2023) The 2022 Political Declaration on the Use of Explosive Weapons in Populated Areas: A tool for protecting the environment in armed conflict? International Review of the Red Cross, 105(924), pp. 1293-1312. Cluster Munition Coalition (2023) Cluster Munition Monitor 2023. Geneva: International Campaign to Ban Landmines - Cluster Munition Coalition. Cluster Munition Coalition (2024) Cluster Munition Monitor 2024. Geneva: International Campaign to Ban Landmines - Cluster Munition Coalition. Cluster Munition Coalition (2025) Cluster Munition Monitor 2025. Geneva: International Campaign to Ban Landmines - Cluster Munition Coalition. Congressional Research Service (2024) Cluster Munitions: Background and Issues for Congress. Report RS22907. Washington DC: Congressional Research Service. Cooper, N. (2022) Humanitarian Disarmament II: Case Studies. In: The Palgrave Encyclopedia of Global Security Studies. Cham: Springer. Docherty, B. (2021) Completing the Package: The Development and Significance of Positive Obligations in Humanitarian Law. In: Dunworth, T. and Hood, A. (eds.) Disarmament Law: Reviewing the Field. Abingdon: Routledge. Docherty, B. and Sanders-Zakre, A. (2023) The origins and influence of victim assistance: Contributions of the Mine Ban Treaty, Convention on the Rights of Persons with Disabilities and Convention on Cluster Munitions. International Review of the Red Cross, 105(922), pp. 252-277. https://doi.org/10.1017/S1816383122000753 Dunworth, T. and Hood, A. (eds.) (2021) Disarmament Law: Reviewing the Field. Abingdon: Routledge. International Committee of the Red Cross (2025) Reneging on the cluster munitions ban endangers civilian lives and erodes international humanitarian law. Geneva: ICRC. Kmentt, A. (2021) The Treaty Prohibiting Nuclear Weapons: How It Was Achieved and Why It Matters. Abingdon: Routledge. LEX International (2024) Lithuania's Withdrawal from the Convention on Cluster Munitions: Briefing Paper. Geneva: LEX International. Mine Action Review (2024) Clearing Cluster Munition Remnants 2024. Oslo: Norwegian People's Aid. Poposka, V. and Nuredin, A. (2024) Legality of the use of cluster bombs in international law: A short overview. Journal of Liberty and International Affairs, 10(1), pp. 252-269. https://doi.org/10.47305/JLIA24101270p Stockholm International Peace Research Institute (2023) Conventional arms control and the regulation of new weapon technologies. In: SIPRI Yearbook 2023: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Stockholm International Peace Research Institute (2024) Conventional arms control and the regulation of inhumane weapons. In: SIPRI Yearbook 2024: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Torelli, C. (2023) Patterns of explosive violence harm in Ukraine: Comparing patterns of harm associated with explosive weapons in Ukraine from 2014 to February 2022, and after February 2022. Disaster Medicine and Public Health Preparedness, 17. United Nations Office for Disarmament Affairs (2024) Convention on Cluster Munitions: Report of the Twelfth Meeting of States Parties. Geneva: UNODA. United Nations Office for Disarmament Affairs (2025) Convention on Cluster Munitions: Report of the Thirteenth Meeting of States Parties. Geneva: UNODA. Weston, N., Cheran, A., Ratnayake, A., Kushner, A. L. and Wild, H. (2025) Context-appropriate clinical practice guidelines for the care of explosive ordnance casualties in low-resource settings. World Journal of Surgery. https://doi.org/10.1002/wjs.12644 Hashtags #Convention_on_Cluster_Munitions #cluster_munitions #humanitarian_disarmament #international_humanitarian_law #victim_assistance #explosive_remnants_of_war #arms_control #protection_of_civilians #Oslo_Process #stockpile_destruction #mine_action #disarmament_law #treaty_compliance #civilian_protection #norm_erosion

The Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on Their Destruction, adopted at Oslo on 18 September 1997 and opened for signature in Ottawa in December 1997, is widely treated as the founding instrument of modern #humanitarian_disarmament. It was the first treaty to ban outright a weapon that was already in wide military use, and it did so by combining a categorical prohibition with a set of positive duties: destroy the stockpiles, clear the ground, help the survivors, and report honestly on progress. This article offers a structured review of the Convention across its first quarter-century and into the turbulent period of 2024 to 2026. It examines the legal architecture of the treaty article by article, evaluates the implementation record against the treaty's own benchmarks, and analyses the most serious crisis the regime has faced: the first withdrawals in its history, together with one State Party's attempt to suspend its obligations while at war. The study uses a doctrinal and documentary method, drawing on the treaty text, the action plans agreed at successive review conferences, the annual reporting of the Landmine Monitor, and recent peer-reviewed scholarship. It finds that the Convention has performed strongly on stockpile destruction and on building a shared professional practice of #mine_action, more modestly on clearance and #victim_assistance, and poorly on insulating itself against a sharp change in the European security environment. The article argues that the current crisis is not evidence that the treaty failed, but evidence of how a #stigma based regime behaves when the perceived cost of compliance rises quickly. It closes with recommendations for States Parties, donors, researchers, and students, and with a discussion of what the episode teaches about the design of prohibition treaties more generally. Keywords: anti-personnel mines; Ottawa Treaty; Mine Ban Treaty; humanitarian disarmament; international humanitarian law; mine action; victim assistance; treaty withdrawal; norm erosion; civilian protection. 1. Introduction Few weapons have attracted the sustained moral attention that #landmines have. A rifle needs a soldier to aim it. A shell needs a gun crew. A mine needs nobody. Once buried, it waits. It does not know when a ceasefire has been signed. It does not know that the farmer walking across the field in 2026 was a child in 1994, and it does not care. This quality, which the technical literature calls being #victim_activated, is the reason that an entire branch of international law grew up around a device that often costs less than a cup of coffee to manufacture and thousands of dollars to remove. The legal answer to that problem is the Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on Their Destruction. It is known by several names, and this article uses them interchangeably: the #Ottawa_Treaty, the #Mine_Ban_Treaty, the Anti-Personnel Mine Ban Convention, or simply the Convention. It was concluded in Oslo on 18 September 1997, opened for signature in Ottawa on 3 and 4 December 1997, and entered into force on 1 March 1999. Its central promise is written into its first operative provision: States Parties undertake never, under any circumstances, to use, develop, produce, acquire, stockpile, retain, or transfer #anti_personnel_mines, and never to assist anyone else in doing so. That promise held, more or less, for twenty-five years. Then it began to crack. Between March 2025 and February 2026, five States Parties on Europe's eastern edge formally left the Convention. Estonia, Latvia, Lithuania, Poland, and Finland each invoked the withdrawal clause, citing the threat posed by Russia and the changed security picture after the full-scale invasion of Ukraine in February 2022. In parallel, Ukraine, which remains a State Party and which cannot lawfully withdraw while it is engaged in an armed conflict, announced in 2025 that it would suspend the operation of the treaty, a move the Convention does not provide for. Meanwhile, the annual casualty figures were moving in the wrong direction: the Landmine Monitor recorded 6,279 people killed or injured by mines and explosive remnants of war in 2024, the highest annual total since 2020, with civilians making up roughly ninety per cent of them and children making up close to half. This is therefore an unusually good moment to study the Convention. A treaty under stress reveals its structure in a way that a treaty at rest does not. When everything is going well, it is hard to tell whether States comply because the law obliges them, because compliance is cheap, or because they never wanted the weapon in the first place. When the cost of compliance rises sharply, the answer becomes visible. 1.1 Research problem The scholarly and policy literature on the Ottawa Treaty tends to fall into two moods. The first is celebratory. It emphasises the #Ottawa_Process, the role of the #International_Campaign_to_Ban_Landmines, the 1997 #Nobel_Peace_Prize, the tens of millions of stockpiled mines destroyed, and the long fall in casualty numbers from the late 1990s. The second mood, which has grown louder since 2022, is elegiac. It emphasises withdrawal, backsliding, funding collapse, and the return of great-power hard security thinking. Neither mood, on its own, explains what is happening. A celebratory reading cannot account for why five European democracies, all of them long-standing supporters of #international_humanitarian_law, decided within a single year that they needed the weapon back. An elegiac reading cannot account for the fact that, in the same period, the Marshall Islands ratified and Tonga acceded, and that the great majority of the world's States remain bound and remain compliant. The research problem addressed here is therefore: what does the record of the Anti-Personnel Mine Ban Convention, including the 2024 to 2026 crisis, tell us about how a humanitarian prohibition regime actually works, where its strength lies, and where its design left it exposed? 1.2 Research questions What is the legal architecture of the Convention, and how do its prohibitive and positive obligations fit together? How has the Convention performed against its own benchmarks in #stockpile_destruction, #mine_clearance, #victim_assistance, #universalization, and transparency? What legal and political mechanisms explain the 2025 to 2026 #withdrawal wave and Ukraine's attempted #suspension? What do these developments imply for the durability of humanitarian prohibition regimes, and what practical steps follow? 1.3 Contribution and significance This article is written for students and for readers who need a single, careful, and honest account of the Convention in one place. Its contribution is threefold. First, it sets out the treaty's legal machinery in plain language without losing legal precision. Second, it evaluates performance against the benchmarks the States Parties set for themselves in their successive action plans, rather than against an abstract ideal. Third, it treats the current crisis as an object of analysis rather than as an occasion for lament, and asks what a treaty designer would have had to do differently, if anything, to prevent it. 1.4 Structure Section 2 reviews the literature. Section 3 explains the method. Section 4 traces the historical background, from the failure of the earlier framework to the Ottawa Process. Section 5 analyses the legal architecture in detail. Section 6 assesses the implementation record. Section 7 identifies structural gaps. Section 8 examines the withdrawal crisis and its legal dimensions. Section 9 discusses the findings in the light of theory. Section 10 offers recommendations, Section 11 states the limitations, and Section 12 concludes. 2. Literature Review 2.1 The first wave: explaining the ban The earliest and largest body of work on the Convention was written by scholars of international relations who wanted to explain something that realist theory had trouble accounting for. A coalition of non-governmental organisations, working with a group of middle powers and without the participation of the United States, Russia, or China, produced a binding treaty in about fourteen months. The classic accounts of this period stress transnational advocacy, the reframing of a military question as a humanitarian one, and the willingness of States such as Canada, Norway, Austria, and South Africa to negotiate outside the consensus-bound machinery of the #Convention_on_Conventional_Weapons. The core insight of this literature is that the ban succeeded because campaigners changed the question. Instead of asking whether mines could be used lawfully if certain technical safeguards were applied, they asked whether a weapon whose predictable long-term effect is to maim farmers and children could be tolerated at all. Once the question was framed that way, the technical debate about self-destructing fuzes and detectability lost much of its force. 2.2 The second wave: implementation and mine action A second and more practical literature grew around the delivery of the treaty's positive obligations. This work is often published in specialised outlets and by operating agencies rather than in mainstream law journals, and it is concerned with survey methods, #land_release standards, cost per square metre, quality management, mechanical and canine detection, and increasingly with drones, remote sensing, and machine learning. Recent contributions from researchers working on Ukraine have expanded this field considerably, because Ukraine presents a contamination problem of a scale and character that the sector has not previously faced (Hutsul et al., 2023; Hutsul, 2025). This literature has an important implicit finding. The binding constraint on clearance is rarely the law. It is money, information, access, and trained people. A State can be entirely willing to comply with its clearance obligation and still miss its deadline by a decade. 2.3 The third wave: victim assistance and disability rights A distinct strand of scholarship examines the treaty's contribution to the idea of #victim_assistance. Docherty and Sanders-Zakre (2023) trace how the Mine Ban Treaty introduced the concept, how the Convention on the Rights of Persons with Disabilities then supplied it with a human-rights framework, and how the Convention on Cluster Munitions later hardened it into a firmer legal obligation. Their analysis matters for this article because it shows that the Ottawa Treaty's most influential contribution may not be its prohibition at all, but its insistence that a disarmament treaty owes something to the people the weapon has already harmed. Related work on accountability and #disability_rights in armed conflict has broadened this discussion further (Pons, Lord and Stein, 2023). 2.4 The fourth wave: humanitarian disarmament under strain The most recent literature, written largely after 2020, is more sceptical. Dunworth (2020) challenges the "big bang" account that presents 1997 as a clean break with an older, State-centred tradition, arguing instead that humanitarian framing has a long and politically complicated history. Cooper (2022) goes further, suggesting that humanitarian disarmament has sometimes produced "disarmament without disarmament", in the sense that treaties are concluded among States that were never the main problem, while the principal users remain outside. Since 2022, a rapidly growing body of policy analysis has focused on the eastern-flank withdrawals. Institutional research, including work produced by the United Nations Institute for Disarmament Research (UNIDIR, 2025) and parliamentary research services (Library of Parliament, 2026), has begun to map the legal and strategic arguments on both sides. 2.5 The gap this article addresses What is missing is an integrated treatment aimed at students that does three things at once: explains the law accurately, reports the empirical record honestly, and analyses the crisis without either defending the withdrawing States uncritically or condemning them without engaging their arguments. This article attempts that integration. 3. Methodology 3.1 Research design This is a qualitative, doctrinal, and documentary study. It does not generate new primary data. It analyses existing legal texts and existing empirical reporting, and it interprets them. Three methods are combined. Doctrinal legal analysis. The treaty text is read closely, article by article, in the light of the general rules of treaty interpretation. Where the text is ambiguous, the article considers the object and purpose of the Convention, subsequent practice of the States Parties as expressed in review conference documents, and the interpretive positions taken by the ICRC and the Convention's own presidencies. Documentary and content analysis. The article draws on the annual reporting of the Landmine Monitor, the research and monitoring arm of the International Campaign to Ban Landmines, on the action plans adopted at successive review conferences, and on the published statements of withdrawing States. Comparative and interpretive analysis. The article compares the Ottawa Treaty's design with that of its successor instruments, particularly the Convention on Cluster Munitions, and uses norm-lifecycle reasoning to interpret the pattern of withdrawal. 3.2 Data sources Sources fall into four categories: Primary legal texts: the Convention itself, and the withdrawal notifications and explanatory statements submitted under it. Institutional monitoring: Landmine Monitor reports, principally the editions from 2022 to 2025; Mine Action Review's clearance reporting; ICRC guidance. Peer-reviewed scholarship: principally the works listed in the References. Official conference outcomes: the action plans agreed at Nairobi (2004), Cartagena (2009), Maputo (2014), Oslo (2019), and the #Siem_Reap_Angkor_Action_Plan agreed at the Fifth Review Conference held in Siem Reap, Cambodia, from 25 to 29 November 2024. 3.3 Analytical framework Performance is assessed against benchmarks that the States Parties themselves adopted, not against benchmarks invented by the author. This matters. It would be easy, and unfair, to judge the Convention against the standard of a mine-free world by any given date. It is more useful to ask whether States did what they promised each other they would do, within the timeframes they set. The framework has five performance dimensions: Prohibition compliance (did States Parties stop using, producing, and transferring?) Stockpile destruction (Article 4) Clearance of mined areas (Article 5) Assistance to victims and international cooperation (Article 6) Transparency and institutional health (Articles 7 to 9, plus the meetings machinery) 3.4 Limitations of method Casualty and contamination data are estimates. They are collected in conditions of insecurity, displacement, and restricted access, and they under-report. This article treats all figures as minimum figures. The article also relies on English-language sources, which may under-represent scholarship published in Russian, Ukrainian, Arabic, Burmese, and Khmer. 4. Historical Background: From Battlefield Tool to Banned Weapon 4.1 What an anti-personnel mine actually is An anti-personnel mine is defined by the Convention as a mine designed to be exploded by the presence, proximity, or contact of a person, and that will incapacitate, injure, or kill one or more persons. Two features of that definition do the work. The first is that the mine is designed to be triggered by the victim. It is not fired by a soldier who makes a judgment about who is in front of him. It is armed and left. Whoever steps on it, triggers it. This is the source of its #indiscriminate_effects: not that the weapon is inaccurate, but that it delegates the decision to kill to the victim. The second is duration. A rifle stops being dangerous when it is put down. A mine does not. Mines laid in Cambodia in the 1980s, in Bosnia in the 1990s, and in Angola across decades continued to kill long after the political reasons for laying them had disappeared. Two broad families exist. Blast mines detonate underfoot and typically destroy the lower limb, often requiring amputation. Fragmentation mines project metal at speed and can injure several people at once. In practice the survivor's experience is similar in both cases: catastrophic injury, long rehabilitation, permanent disability, loss of livelihood, and, very often, poverty. It is important for students to understand what the Convention does not cover. It does not prohibit anti-vehicle mines, which are designed to require the weight of a vehicle. It does not prohibit command-detonated devices, which a person fires deliberately by remote control, because these are not victim-activated. These exclusions were deliberate, and, as Section 7 explains, they have created enduring grey zones. 4.2 The failure of the earlier framework Before 1997, the applicable instrument was Protocol II to the 1980 #Convention_on_Conventional_Weapons. That protocol regulated rather than prohibited. It required mines to be recorded, restricted their use against civilians, and imposed rules about marking and fencing minefields. It did not work. Recording obligations were widely ignored. Minefields were not fenced. Non-state armed groups, which were laying an increasing share of the world's mines, were not bound in any meaningful way. The 1996 revision, known as #Amended_Protocol_II, tightened the rules by requiring mines to be detectable and by encouraging self-destruction and self-deactivation features. It was a real improvement in law. It was also, in the view of the emerging campaign, an admission of defeat, because it accepted that mines would continue to be laid. The campaign's decisive argument was practical rather than philosophical. The technical fixes required disciplined, well-resourced armies with good record-keeping and reliable manufacturing. The mine problem was concentrated in wars fought by forces that had none of these things. A rule that only works if everyone behaves well is not a rule for the world as it is. 4.3 The Ottawa Process By 1996 it was clear that the consensus rule inside the Convention on Conventional Weapons framework would allow any single mine-producing State to block a total ban. Canada's response, announced by its foreign minister at the close of a conference in Ottawa in October 1996, was to invite States that were willing to sign a comprehensive ban to come back within a year and do so. What followed is now studied as a model of fast, coalition-based, civil-society-supported treaty-making. A core group of States, working alongside the #International_Campaign_to_Ban_Landmines and the ICRC, negotiated on the basis of an Austrian draft text, held a diplomatic conference in Oslo in September 1997, and produced the Convention. Attempts to water the text down, including a proposal to break the ban into optional protocols that States could accept selectively, were rejected precisely because they would have preserved the appearance of a ban while removing its substance. In December 1997, 122 governments signed. The treaty entered into force on 1 March 1999, faster than any comparable multilateral instrument of its era, because it required only forty ratifications. Two features of this history matter for what follows. First, the treaty was built by a coalition of the willing, not by the whole international community; the biggest producers and holders stayed out from the start. Second, the treaty's power was always partly moral rather than coercive. It had no inspectorate, no sanctions, and no enforcement body. It relied on transparency, peer pressure, and #stigma. 5. The Legal Architecture of the Convention 5.1 Article 1: the general obligations #Article_1 is short and absolute. Each State Party undertakes never under any circumstances to use anti-personnel mines; to develop, produce, otherwise acquire, stockpile, retain, or transfer them; or to assist, encourage, or induce anyone else to engage in any prohibited activity. Each State Party further undertakes to destroy all anti-personnel mines in accordance with the Convention. The phrase "never under any circumstances" is doing deliberate work. It excludes the defences that States normally reach for. There is no exception for necessity. There is no exception for reciprocity, so a State Party may not use mines merely because its adversary is using them. There is no exception for defensive use as opposed to offensive use. The "assist, encourage or induce" clause has generated real interpretive difficulty in the context of joint military operations. May a State Party take part in a coalition operation with a non-party that uses mines? The generally accepted reading, reflected in the practice of many States Parties, is that mere participation in joint operations is not prohibited, but that active involvement in planning or carrying out mine use is. This is an uncomfortable compromise, and it is one that scholars have criticised as leaving the prohibition thinner than its language suggests. 5.2 Article 2: definitions Article 2 defines the anti-personnel mine, the mine, the mined area, and transfer. Two definitional points are worth close attention. First, the definition of the anti-personnel mine turns on design intent: the mine must be designed to be exploded by the presence, proximity, or contact of a person. This is why a mine designed to be exploded by a vehicle, but which happens to be sensitive enough to be set off by a person, sits in a contested space. Second, the Convention expressly states that mines designed to be detonated by the presence, proximity, or contact of a vehicle, and equipped with anti-handling devices, are not anti-personnel mines by virtue of that anti-handling device. This exclusion was necessary to get some States on board. It also introduced a permanent argument about where the line falls. 5.3 Article 3: the retention exception #Article_3 permits a State Party to retain or transfer a number of anti-personnel mines for the development of, and training in, mine detection, mine clearance, or mine destruction techniques. The number retained must not exceed the minimum number absolutely necessary for those purposes. In practice this narrow exception has become one of the Convention's most persistent compliance irritations. The treaty gives no ceiling, only a standard of absolute necessity, and it requires States to report what they hold. Reporting has been uneven. Some States have held large numbers of #retained_mines for many years without using them for training, which is difficult to reconcile with the words "absolutely necessary". More than sixty States Parties still retain mines under this provision. 5.4 Article 4: stockpile destruction Article 4 requires each State Party to destroy all stockpiled anti-personnel mines it owns or possesses, or that are under its jurisdiction or control, as soon as possible but not later than four years after the Convention enters into force for that State. This is the clearest, most measurable, and most successfully implemented obligation in the treaty. It is measurable because a stockpile is a countable thing in a warehouse. It is achievable because destruction is technically straightforward and comparatively cheap. And it is verifiable in the loose sense that a State which declares a stockpile and then declares its destruction has created a paper trail its peers can examine. More than fifty-five million stockpiled mines have been destroyed since the Convention entered into force. That figure, more than any other, is what supporters of the treaty point to when they are asked what it achieved. 5.5 Article 5: clearance of mined areas #Article_5 requires each State Party to destroy all anti-personnel mines in mined areas under its jurisdiction or control as soon as possible, but not later than ten years after entry into force for that State. Where a State cannot meet the deadline, it may request an extension from the Meeting of the States Parties, which may grant a period of up to ten further years, renewable. The extension mechanism deserves particular attention from students, because it is the hinge on which the treaty's credibility on clearance turns. On one reading it is a sign of weakness. A deadline that can be extended repeatedly is not a deadline. Dozens of States have sought extensions, some more than once, and the aspiration set at the 2014 Maputo Review Conference of a mine-free world by 2025 was plainly not achieved. On another reading, the mechanism is what has kept States inside the tent. A rigid deadline with no escape valve would have converted honest, poor, heavily contaminated States into treaty violators overnight. Instead, the extension procedure converts a potential breach into a managed, transparent, peer-reviewed process. A State requesting an extension must explain what it has done, what remains, why it failed, and what it will now do. That is not nothing. It is a form of #compliance management rather than enforcement. The empirical picture supports the second reading only partially. Clearance has moved slowly. Roughly thirty States Parties still carry clearance obligations, and a small group, including Afghanistan, Bosnia and Herzegovina, Cambodia, Ethiopia, Iraq, Turkiye, and Ukraine, remain massively contaminated. Oman completed clearance in 2025, the first State Party to do so in five years, bringing the number of States Parties that have finished the job since 1999 to around thirty-one. 5.6 Article 6: international cooperation, assistance, and victim assistance Article 6 is the treaty's positive core. It provides that each State Party has the right to seek and receive assistance; that States Parties in a position to do so shall provide assistance for the care, rehabilitation, and social and economic reintegration of mine victims, and for mine awareness programmes; and that they shall cooperate on clearance and stockpile destruction. Three observations follow. First, this was genuinely novel. A disarmament treaty that requires its parties to help the people the weapon has already injured was, in 1997, a new idea. Docherty and Sanders-Zakre (2023) show how this seed grew into a full rights-based obligation in later instruments. Second, the obligation is soft in a specific way. States "in a position to do so" shall provide assistance. There is no formula, no assessed contribution, no burden-sharing key. The result is that #mine_action_funding has always depended on the political will of a small number of donors, which is exactly the vulnerability that surfaced in 2025. Third, victim assistance has consistently been the least well-funded pillar. International support for victim assistance fell by around twenty-three per cent in 2024. This is a hard fact for the treaty's defenders to explain, because it means that the pillar most directly concerned with human dignity has been the one most easily cut. 5.7 Articles 7, 8, and 9: transparency, clarification, and national measures #Article_7 requires annual #transparency_reports covering stockpiles, retained mines, mined areas, clearance progress, destruction programmes, and national implementation measures. This is the Convention's substitute for inspections. It is a system of self-reporting, cross-checked in practice by civil society through the Landmine Monitor. Article 8 provides a procedure for the facilitation and clarification of compliance, including the possibility of a fact-finding mission. It has essentially never been used. States Parties have preferred quiet diplomacy to formal accusation. Article 9 requires each State Party to take all appropriate legal, administrative, and other measures, including the imposition of penal sanctions, to prevent and suppress prohibited activity by persons or on territory under its jurisdiction or control. This is where the ban becomes domestic criminal law. Implementation here has lagged badly: dozens of States Parties have still not reported adequate #national_implementation_measures. 5.8 The institutional machinery The Convention has no organisation in the sense that the Chemical Weapons Convention has the OPCW. It has: Annual #Meetings_of_States_Parties, which review implementation and decide on extension requests. Review Conferences roughly every five years, which adopt action plans: Nairobi (2004), Cartagena (2009), Maputo (2014), Oslo (2019), and Siem Reap-Angkor (2024). An #Implementation_Support_Unit in Geneva, small and funded by States Parties, which provides continuity and technical support. A dense surrounding ecosystem of operators, standard-setters, and monitors, without which the formal machinery would not function. This lean design was a strength as long as compliance was cheap. It offers very little when a State decides to leave. 5.9 Articles 19 and 20: no reservations, and the door marked exit Article 19 provides that the articles of the Convention shall not be subject to reservations. A State cannot join while carving out an exception for itself. This is a deliberately hard-edged provision and it reflects the all-or-nothing logic of the ban. #Article_20 is the provision that has moved from footnote to headline. It provides that the Convention is of unlimited duration, but that each State Party has, in exercising its national sovereignty, the right to withdraw. Withdrawal requires notification to all other States Parties, to the Depositary (the UN Secretary-General), and to the UN Security Council, and that notification must include a full explanation of the reasons motivating the withdrawal. Two conditions then apply. The withdrawal takes effect six months after receipt of the notification. And, critically, if on the expiry of that six-month period the withdrawing State Party is engaged in an #armed_conflict, the withdrawal does not take effect before the end of that conflict. That last clause was borrowed from the tradition of the Geneva Conventions. Its purpose is to prevent what has been called opportunistic renunciation: a State fighting a war cannot simply exit the rules in the middle of it. As Section 8 shows, this clause is now the single most consequential sentence in the entire treaty. 6. The Implementation Record, 1999 to 2026 6.1 Prohibition compliance For most of the treaty's life, compliance with the core prohibition by States Parties was close to complete. Global production collapsed: the number of States producing anti-personnel mines fell from more than fifty to fewer than ten. The legal trade effectively ended; what remains is illicit and small. Use by States Parties became rare enough that each instance was treated as a scandal. That picture has now changed at the margins, and the margin is where the danger lies. The Landmine Monitor reports continued and extensive use by States not party, principally Russia and Myanmar. It also reports indications of new use by Ukraine, a State Party, including devices apparently delivered by drone, and allegations of use by Cambodian forces along the Thai border. Non-state armed groups used mines, mostly improvised, in at least ten States Parties. 6.2 Stockpile destruction This is the treaty's clearest success. Over fifty-five million stockpiled mines have been destroyed. The overwhelming majority of States Parties met their four-year deadline. A small number of laggards remain, and their non-compliance is well documented, but it is a residual problem rather than a systemic one. The lesson for treaty design is worth stating plainly. Obligations that are countable, time-bound, technically simple, and cheap get done. Obligations that are open-ended, expensive, and dependent on other people's money do not. 6.3 Clearance Clearance is the hard case. Contamination affects at least fifty-seven States and other areas, including around thirty-two States Parties. Around thirty-one States Parties have completed clearance since 1999. Progress is real but slow, and it is being outrun by new contamination. Ukraine alone has transformed the global picture. Estimates of the cost of clearing Ukraine run to tens of billions of United States dollars over a decade or more, and estimates of the annual economic loss caused by contamination run to billions per year, concentrated in agriculture, energy, and transport (Hutsul et al., 2023). The 2025 aspiration was not met. Practitioners now speak, with some resignation, of 2030 as the new 2025, and even that assumes resources that do not currently exist. 6.4 Casualties The long-run trend is genuinely encouraging and the short-run trend is genuinely alarming, and both statements are true at once. In the late 1990s, annual casualties were estimated at around twenty-five thousand. By the mid-2010s the annual figure had fallen to a few thousand. The Convention, together with the mine action sector it helped create, deserves substantial credit for that fall. But the figures have been climbing again. The Landmine Monitor recorded 6,279 #casualties in 2024, the highest since 2020 and an increase of about nine per cent on the previous year. Civilians accounted for roughly ninety per cent. #children accounted for almost half of all casualties, and in Afghanistan more than three quarters. Myanmar recorded the highest national total, followed by Syria, Afghanistan, Ukraine, and Yemen. Preliminary data for 2025 indicate more than five thousand casualties across the countries reviewed, with Syria recording the highest figure. Behind each of these numbers is a person, and behind each person is a household that has usually lost its main earner or its main carer. 6.5 Victim assistance Assistance to survivors has advanced conceptually and stalled financially. Conceptually, the Convention pushed the sector to stop treating survivors as a special category of beneficiary and to integrate them into national disability, health, and social protection systems, in line with the framework of the Convention on the Rights of Persons with Disabilities (Docherty and Sanders-Zakre, 2023; Pons, Lord and Stein, 2023). Financially, the picture is poor. Victim assistance funding fell sharply in 2024, and health systems in conflict-affected countries are struggling with a rising number of amputations. Survivors in rural areas remain the least served, and lifetime costs of prosthetics, revision surgery, and psychosocial support are chronically underestimated. 6.6 Universalization and transparency Membership peaked at around 166 States Parties. Following the withdrawals discussed below, the number stands at roughly 161 as of early 2026. In the same period the Marshall Islands ratified and Tonga acceded, which is a reminder that #universalization has not stopped, even as the regime lost members at the other end. Transparency reporting remains reasonably healthy, though not universal. Late and incomplete Article 7 reports are a recurring complaint at Meetings of the States Parties, and the retention figures reported under Article 3 continue to be difficult to reconcile with the treaty's language. 6.7 Funding International support for mine action was around 761 million United States dollars in 2024, a decline of about five per cent on the previous year. In early 2025 the United States, historically the largest single donor, imposed a sector-wide freeze on foreign assistance, followed by the termination of several humanitarian mine action programmes. Programmes in Afghanistan, Iraq, Yemen, Colombia, Tajikistan, and Zimbabwe were shut down. This is the treaty's structural weakness made visible. A regime whose positive obligations depend on voluntary contributions from a handful of donors is one budget decision away from a crisis it cannot control. #donor_dependence is not a footnote to the story; it is a central part of it. 7. Structural Gaps and Persistent Challenges 7.1 The improvised mine problem The Convention was written with factory-made mines in mind. The weapon that now causes the most harm in many contexts is the improvised anti-personnel mine, built from fertiliser, scrap, and salvaged explosive, and laid by #non_state_armed_groups. Legally, the position is clearer than is sometimes assumed. If a device is victim-activated and designed to injure or kill a person, it is an anti-personnel mine within the meaning of Article 2 regardless of where it was made. The Convention's States Parties have confirmed this reading. #improvised_mines are covered. Practically, the position is much harder. Non-state armed groups are not parties to the treaty and cannot become parties. The obligation therefore falls on the territorial State to suppress the activity under Article 9, which is precisely the State least able to do so, because if it controlled the territory the groups would not be laying mines there. 7.2 The 2025 deadline that was missed The Maputo commitment to complete clearance by 2025 was aspirational rather than legally binding, but it was a public promise, and it was not kept. The failure has a cost beyond the ground that remains uncleared: it weakens the argument that patience pays. When a State Party is told to keep faith with a slow process, it is reasonable for it to ask what the last set of promises delivered. 7.3 The definitional grey zones Two grey zones persist. The first concerns anti-vehicle mines fitted with fuzes so sensitive that a person can detonate them. Formally these fall outside the Convention because of their design intent; in the field, the distinction can be invisible to the victim. The second concerns command-detonated devices, which are excluded because a human being makes the firing decision. Where the "command" is delivered by an automated sensor, or by a remote operator watching a video feed, the line between a command-detonated device and a victim-activated one becomes conceptually thin. Emerging drone-delivered munitions sharpen this problem further. 7.4 Retention under Article 3 The absence of a numerical ceiling in Article 3 has produced two decades of unresolved argument. States Parties have repeatedly urged one another to reduce retained holdings, review whether they are genuinely necessary, and report on their actual use in training. Progress has been slow, and the credibility cost is real, because a treaty that cannot police its own narrow exception invites questions about the rest. 7.5 National implementation legislation Dozens of States Parties have never enacted the penal legislation required by Article 9. In practice this means that, in those States, the ban exists in international law but not in the domestic criminal code, and no individual could be prosecuted for violating it. For a #compliance system that relies on domestic implementation rather than international enforcement, this is a serious hole. 8. The Crisis of 2024 to 2026: Withdrawal, Suspension, and Norm Erosion 8.1 What happened In July 2023, Eritrea, which had been found non-compliant for failing to report on clearance, announced an intention to withdraw and then rescinded the notice within months. That episode was treated at the time as an oddity. It was not. On 18 March 2025, the defence ministers of Estonia, Latvia, Lithuania, and Poland jointly recommended withdrawal from the Convention, citing the deterioration of the security environment on NATO's eastern border. On 1 April 2025, the Finnish government announced the same intention; the Finnish parliament voted for withdrawal on 19 June 2025, the instrument was deposited on 10 July 2025, and the withdrawal took effect on 10 January 2026. The three Baltic States completed their withdrawals in December 2025 and #Poland in February 2026. Separately, Ukraine, which cannot lawfully withdraw while the war continues, announced in 2025 that it would suspend the operation of the Convention. Within roughly one year, therefore, the regime lost five members and saw a sixth attempt to step outside its obligations without leaving. 8.2 The legal analysis of withdrawal It is important to be precise. The withdrawals by Estonia, Latvia, Lithuania, Poland, and Finland were lawful. Article 20 permits withdrawal. Those States gave notice, gave reasons, and waited out the six-month period. None of them was engaged in an armed conflict at the relevant time, so the suspensive clause in Article 20 did not bite. This produces a conclusion that students often find uncomfortable, and it is worth sitting with. A State can comply perfectly with the law of the treaty while destroying the purpose of the treaty. Legality and fidelity are not the same thing. The Convention was drafted by people who understood this, which is why they insisted on a full explanation of reasons and on notification to the Security Council: not to prevent withdrawal, but to make it politically expensive. The strategic timing is also revealing. Analysts have pointed out that these States acted while they still lawfully could. Once a State is in an armed conflict, the exit closes. The eastern-flank withdrawals can therefore be read as insurance policies taken out before the storm, and this is exactly the incentive structure that Article 20 creates. A clause designed to prevent opportunistic exit during war may, in practice, encourage pre-emptive exit before war. 8.3 Ukraine's attempted suspension Ukraine's position is legally different and, in the view of most commentators, weaker. The Convention contains no suspension clause. Ukraine cannot withdraw, because Article 20 postpones the effect of any withdrawal until the end of an armed conflict in which the withdrawing State is engaged. Ukraine therefore sought to rely on general rules of the law of treaties, arguing that circumstances had fundamentally changed since it ratified. The difficulties with this are considerable. The relevant general rules, as reflected in the #Vienna_Convention_on_the_Law_of_Treaties, treat fundamental change of circumstances as an exceptional ground, and they specifically exclude reliance on it where the change results from a breach by the party invoking it, and where the treaty's obligations were assumed precisely with the contingency of war in mind. A treaty about the conduct of hostilities is, almost by definition, a treaty whose obligations are meant to apply when hostilities occur. The argument that war is an unforeseen circumstance justifying suspension of a law-of-war treaty is a hard one to sustain. The ICRC, the ICBL, and other actors have accordingly taken the position that the Convention does not permit suspension of obligations, including in wartime, and that Ukraine remains bound. None of this diminishes the difficulty of Ukraine's position. It is a State Party being invaded by a non-party that uses mines extensively and at scale. The legal answer and the strategic dilemma point in different directions. Recognising that tension honestly is more useful than pretending it does not exist. 8.4 The case made by the withdrawing States Intellectual honesty requires setting out the arguments of the withdrawing States as they themselves would make them, and this section does so without endorsing them. #military_necessity and asymmetry. The withdrawing States argue that they face an adversary that is not bound by the Convention, that uses mines freely, and that has demonstrated a willingness to invade a neighbour. In their view, a treaty that binds the defender and not the attacker creates an asymmetry that a small State on a long border cannot afford. #area_denial and #deterrence. They argue that mines are among the few cost-effective ways for a numerically inferior force to slow an armoured advance, canalise it into prepared killing zones, and buy the time needed for allies to arrive. They point to the role that Russian minefields played in blunting Ukraine's 2023 counteroffensive as evidence that mines retain battlefield relevance. Sovereignty and democratic decision. They argue that Article 20 exists precisely so that a State whose security circumstances change can revisit its commitment, and that these decisions were taken by elected parliaments after public debate, not by executive fiat. Mitigation. Several have indicated that any future use would be accompanied by mapping, marking, and post-conflict clearance commitments, and that they would remain donors to mine action. 8.5 The case against The counter-arguments are equally worth stating. The weapon does not distinguish, and the mitigation promises are historically unreliable. Every State that has laid mines has promised to record them. Records are lost, units are overrun, maps are wrong. Bosnia and Herzegovina, thirty years after the fighting stopped, is still contaminated. The cost falls on the withdrawing State's own citizens. Mines laid on Polish, Baltic, or Finnish soil would contaminate Polish, Baltic, and Finnish soil. Clearance costs and civilian casualties would be borne domestically for decades. The Ukrainian experience, where clearance is estimated to require tens of billions of dollars and a generation of work, is the relevant precedent, not a distant one. Military utility is contested. Critics argue that mines are a tactical, not a strategic, asset; that modern sensors, loitering munitions, and precision fires have substantially displaced the area-denial function; and that mines are ineffective against the drone-heavy, dispersed, infantry-light tactics that now characterise the fighting in Ukraine. Normative contagion. This is the argument that carries the most weight for international lawyers. When five European democracies, which for decades presented themselves as guardians of #international_humanitarian_law, leave a humanitarian treaty citing security, they hand a ready-made script to every other State that finds a humanitarian obligation inconvenient. The UN High Commissioner for Human Rights and the ICRC have both warned in these terms: adhering to humanitarian treaties in peacetime and abandoning them in wartime hollows out the entire framework, because the framework exists for wartime. 8.6 Norm erosion: a theoretical reading It is tempting to describe what has happened as the reversal of a norm cascade. That is too neat. A more careful reading is that the anti-personnel mine taboo was always conditional in a way that its supporters preferred not to notice. Between roughly 1997 and 2012, the perceived military value of mines to most States Parties was low. Renouncing them cost little. The moral case was strong, the practical case was cheap, and the norm looked robust because it was never tested. The eastern-flank withdrawals are the test. When the perceived security value of the weapon rose sharply, a group of States revised their assessment. What this shows is not that the #stigma was fake, but that stigma alone is a thin foundation once the cost of compliance rises. Cooper's (2022) suspicion that humanitarian disarmament sometimes amounts to disarmament without disarmament, achieved among States that never really needed the weapon, receives uncomfortable support here. Two qualifications should temper any conclusion of collapse. First, the numbers. Five States left, out of more than a hundred and sixty. Two joined. The regime is bruised, not broken. Second, the direction of the argument. Even the withdrawing States did not argue that mines are humane, or that the humanitarian evidence was wrong. They argued that they had no choice. That is a concession, not a rejection. A norm that forces its violators to plead necessity rather than deny wrongdoing is still doing work, and this is the ground on which the regime can be rebuilt. 9. Discussion 9.1 What the Convention got right A categorical rule. The refusal to permit reservations, and the refusal to accept a regulatory rather than prohibitory model, gave the treaty a clarity that regulatory instruments lack. There is no arguing about whether a particular use was proportionate. Use is use. Positive obligations. By requiring clearance and victim assistance, the treaty gave affected States a reason to join beyond abstract virtue. It converted the treaty from a renunciation into a partnership, and it created the professional field of #mine_action. A workable compliance culture. In place of inspections, the treaty built a culture of annual reporting, peer review, and independent monitoring by civil society. The Landmine Monitor is, in effect, a non-governmental verification body that no treaty negotiator designed and that no State controls. That is an underappreciated institutional achievement. A template. The Convention's architecture was reused in the Convention on Cluster Munitions and, in modified form, in the Treaty on the Prohibition of Nuclear Weapons. It demonstrated that humanitarian treaties can succeed without the participation of the major powers. 9.2 What the Convention got wrong, or left undone No funding mechanism. The single greatest design weakness. Positive obligations were created without an assessed, predictable means of paying for them. The 2025 funding shock exposed this immediately. A soft exit. Article 20 was drafted on the assumption that withdrawal would be politically unthinkable. That assumption held for twenty-five years and then failed. There was no requirement of a waiting period longer than six months, no requirement to consult the other States Parties before deciding, and no procedural cost beyond explanation. An extension mechanism without teeth. The Article 5 extension procedure kept States inside the treaty but allowed clearance deadlines to become notional. Silence on the money side of victim assistance. The obligation to assist victims is real but unpriced. The result is that it is always the first line cut. 9.3 The deeper lesson for treaty design The Ottawa Treaty is often taught as a story about the power of moral argument. It is that. But the fuller lesson is about the interaction between moral argument and perceived cost. Prohibition regimes are strongest where three conditions hold together: the humanitarian harm is vivid and well documented; the military utility of the weapon is genuinely low or declining; and the obligations imposed are cheap, countable, and quick. Where all three hold, as with stockpile destruction, compliance is close to total. Where the third condition fails, as with clearance and victim assistance, progress is slow. And where the second condition reverses, as it did on Europe's eastern flank after February 2022, the first condition on its own turns out to be insufficient. For students of international law, that is the most important sentence in this article. Moral stigma is necessary but not sufficient. It must be reinforced by interest, by institutions, and by money. 10. Policy Recommendations 10.1 For States Parties Reaffirm and formalise the interpretation of Article 20. States Parties should collectively record their understanding that withdrawal is legally available but normatively exceptional, that the "full explanation of reasons" requirement is substantive rather than formal, and that a withdrawing State remains bound by its clearance and victim assistance obligations in respect of contamination it created while a party. Reject suspension. The Meetings of the States Parties should continue to state clearly that the Convention admits no suspension of obligations, including during armed conflict. Keep the door open. Withdrawal should be treated as a suspension of membership in a community, not an excommunication. The regime's long-term interest lies in the return of the withdrawing States, and the diplomatic tone should reflect that. Close the Article 3 loophole. Agree a numerical ceiling for #retained_mines, or a strict use-it-or-destroy-it reporting rule. Finish the domestic legislation. The forty-plus States Parties without adequate penal legislation should be given technical assistance and a deadline. 10.2 For the funding architecture Diversify away from #donor_dependence. The 2025 freeze demonstrated that a sector reliant on one dominant donor is one policy change away from collapse. Pooled funds, regional financing, affected-State budget lines, and innovative financing instruments all deserve serious exploration. Ring-fence victim assistance. Because it is the pillar most easily cut, it is the pillar that most needs a protected budget line. Link mine action to development financing. Clearance is not charity; it is a precondition for agriculture, transport, energy, and the achievement of several #Sustainable_Development_Goals. Framing it that way opens funding channels that humanitarian appeals cannot reach. 10.3 For operators and researchers Invest in evidence about military utility. The withdrawal debate turned on a claim, largely untested in open literature, that mines retain decisive defensive value. Rigorous, publicly available analysis of the actual battlefield performance of mines in contemporary drone-saturated warfare would materially change this debate in either direction. Accelerate #demining_technology. Remote sensing, uncrewed aerial systems, machine-assisted data analysis, and improved information management have begun to change clearance productivity (Hutsul, 2025). Scaling this is the fastest available route to closing the gap between contamination and clearance. Improve casualty data. All reported figures are minimums. Better surveillance, particularly in active conflict zones, would strengthen every argument the regime makes. 10.4 For students and educators Study the treaty as a system, not a slogan. The Convention is not simply a moral victory or a cautionary tale. It is a working system with measurable outputs, identifiable design flaws, and a live stress test running in real time. That makes it one of the best available case studies in international law. 11. Limitations This article has several limitations that readers should hold in mind. It is a documentary study, not a field study. It reports what monitoring bodies and scholars have found; it does not independently verify contamination, casualty, or clearance figures. Its data are lagged. Casualty figures for a given year are published roughly a year later and are revised afterwards. The 2024 figures cited here will be adjusted upward, as such figures usually are. Its coverage of the withdrawal crisis is necessarily provisional. The events described were still unfolding at the time of writing, and the legal consequences, particularly for Ukraine's position, have not yet been authoritatively settled by any court or by the States Parties collectively. Finally, the article relies on English-language sources, and it does not draw on the internal military assessments that would be needed to evaluate the military-utility claims made by the withdrawing States. Those assessments are classified, and their absence from the public debate is itself part of the problem. 12. Conclusion The Anti-Personnel Mine Ban Convention was an extraordinary achievement and it remains one. It banned a weapon that was in wide use, it did so quickly, it did so without the great powers, and it delivered results that can be counted: more than fifty-five million stockpiled mines destroyed, production reduced from more than fifty producer States to fewer than ten, roughly thirty-one States Parties finished with clearance, and a fall in annual casualties from around twenty-five thousand in the late 1990s to a few thousand today. It also has real failures. Clearance is decades behind schedule. The 2025 goal was missed. Victim assistance is chronically underfunded and was cut by nearly a quarter in a single year. Dozens of States Parties have never passed the domestic laws the treaty requires. And in 2025 and 2026, for the first time, States began to walk out. The temptation is to choose one of these stories. The more useful conclusion is that both are true, and that the second follows from a weakness in the first. The Convention rested on a moral stigma that was strong precisely because, for most of its members for most of its life, obeying it was cheap. When obedience stopped being cheap for a handful of States on a militarised border, the stigma alone did not hold them. That is not a reason to abandon the project. It is a reason to understand it more precisely. The five States that left did not argue that mines are acceptable weapons. They argued that they had no alternative. The task facing the remaining States Parties, the ICRC, the campaigners, the operators, and the researchers is to make that claim false: to fund clearance properly, to fund survivors properly, to produce the evidence about military utility that the debate is currently missing, and to hold open the door through which the departed States might one day walk back. A mine laid today will still be there in 2056. Whatever is decided now will be dug out of the ground, one square metre at a time, by someone who is currently a child. That is the timescale on which this treaty should be judged, and it is the timescale on which its defenders should be prepared to work. Hashtags #Ottawa_Treaty_1997 #Anti_Personnel_Mine_Ban_Convention #MineBanTreaty #LandmineFreeWorld #HumanitarianDisarmament #InternationalHumanitarianLaw #MineAction #DeminingMatters #VictimAssistanceNow #ProtectCiviliansInWar #LandmineMonitor #StopLandmines #DisarmamentLaw #TreatyWithdrawalCrisis #ArmsControlStudies References Cooper, N. (2022). Humanitarian disarmament and the era of disarmament without disarmament. In O. P. Richmond and G. Visoka (Eds.), The Palgrave Encyclopedia of Peace and Conflict Studies. Cham: Palgrave Macmillan. https://doi.org/10.1007/978-3-030-77954-2_58 Docherty, B. and Sanders-Zakre, A. (2023). The origins and influence of victim assistance: Contributions of the Mine Ban Treaty, Convention on the Rights of Persons with Disabilities and Convention on Cluster Munitions. International Review of the Red Cross, 105(922), 252-277. https://doi.org/10.1017/S1816383122000753 Dunworth, T. (2020). Humanitarian Disarmament: An Historical Enquiry. Cambridge: Cambridge University Press. Hutsul, T. (2025). Geographic information systems in humanitarian demining: History, development, and modernity. Bulletin of Taras Shevchenko National University of Kyiv: Military-Special Sciences, 3(63), 85-92. https://doi.org/10.17721/1728-2217.2025.63.85-92 Hutsul, T., Myronchuk, K., Tkach, V. and Khobzei, M. (2023). Economic efficiency and priority of demining: International experience. Ukrainian Journal of Applied Economics and Technology, 8(2). https://doi.org/10.36887/2415-8453-2023-2-44 International Campaign to Ban Landmines - Cluster Munition Coalition. (2022). Landmine Monitor 2022. Geneva: ICBL-CMC. International Campaign to Ban Landmines - Cluster Munition Coalition. (2023). Landmine Monitor 2023. Geneva: ICBL-CMC. International Campaign to Ban Landmines - Cluster Munition Coalition. (2024). Landmine Monitor 2024. Geneva: ICBL-CMC. International Campaign to Ban Landmines - Cluster Munition Coalition. (2025). Landmine Monitor 2025. Geneva: ICBL-CMC. International Committee of the Red Cross. (2026). Frequently Asked Questions: The Anti-Personnel Mine Ban Convention (Ottawa Treaty). Geneva: ICRC. Library of Parliament. (2026). The Future of the Ottawa Treaty (Publication No. 2025-20-E). Ottawa: Library of Parliament, Parliamentary Information, Education and Research Services. Mine Action Review. (2024). Clearing the Mines 2024. Oslo and Geneva: Norwegian People's Aid and Mine Action Review. Mine Action Review. (2025). Clearing the Mines 2025. Oslo and Geneva: Norwegian People's Aid and Mine Action Review. Pons, W. I., Lord, J. E. and Stein, M. A. (2023). Addressing the accountability void: War crimes against persons with disabilities. International Review of the Red Cross, 105(922), 465-483. United Nations Institute for Disarmament Research. (2025). New Directions for the Anti-Personnel Mine Ban Convention. Geneva: UNIDIR. United Nations Office for Disarmament Affairs. (2024). Fifth Review Conference of the States Parties to the Anti-Personnel Mine Ban Convention: Siem Reap-Angkor Action Plan 2025-2029. Geneva: UNODA. United Nations Office for Disarmament Affairs. (2025). Anti-Personnel Landmines Convention: Status of the Convention and Implementation Overview. Geneva: UNODA. United Nations Mine Action Service. (2025). Secretary-General's Campaign on Humanitarian Disarmament and Mine Action. New York: UNMAS.

The Treaty on the Prohibition of Nuclear Weapons, adopted at the United Nations on 7 July 2017 and in force since 22 January 2021, is the first #legally_binding international agreement to prohibit nuclear weapons in a comprehensive way. Chemical weapons, biological weapons, anti-personnel landmines and cluster munitions were all outlawed by category before 2017. Nuclear weapons, the most destructive of all, were not. This article explains how that gap was closed. It traces the political and legal history that produced the treaty, sets out its structure article by article in plain language, and assesses what the instrument does and does not achieve. The study uses a qualitative doctrinal and policy analysis based on the treaty text, the records of the three Meetings of States Parties held in 2022, 2023 and 2025, and peer reviewed scholarship published between 2021 and 2026. Four findings emerge. First, the treaty is best understood as the legal product of the #humanitarian_initiative, a decade of diplomacy that reframed nuclear weapons as a question of human consequences rather than of strategic balance. Second, its prohibitions are broader than those of any earlier nuclear instrument and include the threat of use, a provision that speaks directly to the doctrine of #nuclear_deterrence. Third, its positive obligations on #victim_assistance and #environmental_remediation introduce a justice dimension that no previous nuclear agreement contained. Fourth, its practical effect is constrained by the absence of every nuclear armed state and every state that shelters under an extended nuclear guarantee, which means the treaty operates for now as an instrument of norm construction rather than of physical disarmament. The article concludes that the treaty has already changed the language and the moral terms of nuclear politics, that its long term influence depends on #universalization and on the credibility of its verification design, and that the first Review Conference scheduled for late 2026 will be an important test of whether the ban can convert normative weight into material change. Keywords: nuclear disarmament; international law; humanitarian disarmament; nuclear deterrence; treaty law; global nuclear order; arms control; norm diffusion 1. Introduction For more than seventy years, nuclear weapons occupied a strange position in international law. They were the only category of weapon of mass destruction that no universal #treaty_law instrument prohibited outright. #biological_weapons were banned in 1972. #chemical_weapons were banned in 1993. #landmines followed in 1997 and #cluster_munitions in 2008. The weapon capable of destroying entire cities in a single flash, and of altering the climate of the whole planet if used on a large scale, remained lawful to possess for a small group of states and was governed only by rules about who could have it, not about whether anyone should. The Treaty on the Prohibition of Nuclear Weapons closed that gap. It was adopted on 7 July 2017 by a #vote of 122 states in favour, one against and one abstention. It opened for signature on 20 September 2017. #entry_into_force followed on 22 January 2021, ninety days after Honduras deposited the fiftieth instrument of ratification in October 2020. For its parties, it forbids the development, testing, production, manufacture, acquisition, possession, stockpiling, transfer, stationing, use and threat of use of #nuclear_weapons, along with any assistance, encouragement or inducement of those acts. This is a short list of verbs, but it carries a large meaning. Every earlier nuclear instrument accepted, at least implicitly, that some states could lawfully hold these weapons. The Treaty on the Non-Proliferation of Nuclear Weapons of 1968 divided the world into five recognised nuclear weapon states and everyone else, and asked the five to negotiate in good faith towards disarmament at some unspecified point. The #CTBT of 1996 restricted #nuclear_testing but never entered into force. Regional #nuclear_weapon_free_zones banned the weapons across large parts of the southern hemisphere but left the arsenals of the great powers untouched. The 2017 treaty is different in kind. It says that nuclear weapons are unlawful as such for those who accept the obligation, and it invites the rest of the world to join. The purpose of this article is to give students and early career researchers a clear, accurate and reasonably complete account of that instrument. Nuclear policy writing is often either technical to the point of impenetrability or advocacy dressed as analysis. This article tries to avoid both. It explains what the treaty says, why it was written, what its supporters expect from it, what its critics say against it, and what the evidence so far suggests. Three research questions guide the study. What political and legal processes produced a comprehensive prohibition on nuclear weapons after decades of stalemate in traditional #arms_control forums? What legal obligations does the treaty actually create, and how do they differ from those of the existing nuclear regime? What measurable effects has the treaty had since entry into force, and what obstacles stand in the way of further progress? The article proceeds in fourteen sections. Section 2 reviews the scholarly literature. Section 3 sets out the method. Sections 4 and 5 cover the historical background and the negotiation. Sections 6 and 7 examine the legal architecture, including the positive obligations that distinguish the treaty from earlier agreements. Section 8 describes the institutional life of the treaty since 2021. Section 9 analyses membership patterns. Section 10 addresses the relationship with the non-proliferation regime. Section 11 presents the case made by the treaty's opponents. Section 12 assesses normative and practical impact. Section 13 discusses obstacles and pathways. Section 14 notes the limits of this study, and Section 15 concludes with recommendations. 2. Literature Review Scholarship on the treaty has grown quickly since 2021 and can be sorted into five broad strands. 2.1 Insider and historical accounts The most detailed narrative of the treaty's origins comes from Kmentt (2021), who was Austria's lead disarmament diplomat and one of the principal architects of the process that produced it. His account is valuable precisely because it is partisan in a declared way. He reconstructs the internal strategy of the #humanitarian_initiative, the conferences at Oslo, Nayarit and Vienna, and the calculations that convinced a coalition of non-nuclear states to abandon consensus forums and legislate without the great powers. Acheson (2021) writes from within #civil_society and links the ban movement to feminist and anti-colonial critiques of the nuclear order, arguing that the renewed focus on humanitarian and environmental harm was a deliberate attempt to devalue and delegitimise the weapons themselves. 2.2 Structural and ideological critique Egeland (2021, 2022, 2024) supplies the most theoretically ambitious body of work. He argues that the nuclear order is sustained not only by material power but by an ideology that presents disarmament as a distant and slightly unserious aspiration while treating deterrence as sober realism. His book on the long struggle for abolition places the 2017 treaty in a much older lineage of disarmament diplomacy and warns that a narrow focus on #stigmatization can obscure the parallel work that nuclear armed states do to relegitimise their arsenals. Egeland and Pelopidas (2025) extend this critique to the research field itself, documenting how heavily nuclear policy analysis depends on funding from actors with an interest in the status quo. Students should read this literature carefully: it is a reminder that the debate is not a neutral contest between equally resourced positions. 2.3 Quantitative and behavioural studies A second cluster asks who joins the treaty and why. Onderco and Vignoli (2024) applied text-as-data methods to the national statements delivered at the first two Meetings of States Parties and found that participants can be placed along a single axis running from a security framing of disarmament to a humanitarian and emancipatory framing. Membership of a nuclear weapon free zone, especially in Latin America and Africa, has a statistically significant effect on where a state sits. Importantly, they show that the parties are not a single bloc: there are real internal differences about what the treaty is for. Their related work on ratification timing (Onderco and Vignoli, 2025) finds that states are slower to join treaties that impose heavy obligations and that involvement in rivalries delays ratification, which helps explain the geography of membership. Vignoli and Onderco (2025) add that the ideology of the government in office affects commitment to multilateral treaties, and Portela, Onderco and Vignoli (2025) examine how European coordination shapes voting on nuclear questions at the General Assembly. 2.4 Global South and developmental perspectives Futter and Samuel (2024) make an argument that is often missed in Western commentary. The treaty, they show, could not have been adopted without the support of Non-Aligned and Global South states, and that support was conditional on protecting the right to peaceful nuclear technology. They call the resulting worldview "Nutopia": a belief in both the dystopian potential of nuclear weapons and the utopian promise of nuclear energy. The preamble's careful language on the inalienable right to peaceful uses is therefore not decoration. It was the price of the majority. Rajagukguk and Siagian (2024) provide a complementary regional survey of ratification patterns, and Soeparna (2023) links the prohibition to a broader human security agenda. 2.5 The resistance of the nuclear armed states and their allies Gibbons (2022, 2024) analyses how great powers respond to emerging norms they dislike. Her work on United States leadership in the non-proliferation regime shows how much of the regime's stability has depended on active American persuasion, and her later article treats opposition to the ban treaty as a case of deliberate norm resistance rather than passive indifference. Tannenwald (2024) argues that the non-proliferation regime is now caught in self-undermining dynamics: the failure of the nuclear weapon states to deliver on disarmament corrodes the legitimacy of the bargain that non-proliferation depends on. On the allied side, Nishida (2024) offers an unusually candid analysis of Japan, the only country to have suffered nuclear attack and yet a state that has declined even observer status at the Meetings of States Parties. Kanno and Miwa (2024) test Japanese public opinion and find that both external military threats and international pressure shape support for the treaty. Hummel (2024) examines the Belgian parliamentary debate over #nuclear_sharing, and Mukai (2025) presses Japan to move from vague sympathy to concrete commitment. 2.6 The gap in the literature Three gaps are visible. First, most empirical work stops at the second Meeting of States Parties in 2023 and has not yet absorbed the decisions of the third meeting in March 2025 or the preparations for the first Review Conference. Second, there is very little rigorous work on implementation: what states parties have actually done at the national level under Article 5. Third, the literature on Articles 6 and 7 remains dominated by advocacy documents rather than independent evaluation. This article cannot fill those gaps, but it maps them. 3. Research Methodology 3.1 Design This is a qualitative study combining doctrinal legal analysis with policy analysis. Doctrinal analysis means reading the treaty as lawyers read it: examining the ordinary meaning of its terms, its structure, the relationship between its provisions, and the interpretive decisions taken by its parties. Policy analysis means asking what the instrument does in the world: who joins it, who resists it, and what changes as a result. 3.2 Sources Three source types were used. Primary legal materials. The treaty text itself, together with the reports and declarations of the first Meeting of States Parties (Vienna, June 2022), the second (New York, November and December 2023) and the third (New York, March 2025), and the associated action plans and working group outputs. These documents establish how the parties themselves interpret their obligations, which is the most authoritative guide available in the absence of judicial rulings. Peer reviewed scholarship. Journal articles and academic books published between 2021 and 2026 were prioritised, in line with the aim of presenting the current state of knowledge. Older works are referred to in the text where they are historically necessary but are not relied on for present day claims. Aggregate data on membership and arsenals. Figures on ratification and on warhead numbers are drawn from the standard annual compilations of the field. 3.3 Analytical framework The article uses a norm life cycle framework familiar from constructivist international relations, adapted for legal analysis. A norm is understood to pass through emergence, cascade and internalisation, a sequence often described as #norm_diffusion. The framework is used descriptively rather than predictively. It helps organise the evidence, but the article does not assume that the prohibition norm will inevitably complete the cycle. That assumption would beg the central question. 3.4 Limits of the method Doctrinal analysis cannot measure behavioural change, and no state party to the treaty possesses nuclear weapons, so compliance with the core prohibition is trivially easy to observe and tells us nothing. The interesting questions concern indirect effects on non-parties, and these are hard to isolate. Where causal claims are made in this article, they are made cautiously and with the counter-arguments stated. 4. Historical Background: From the Ashes to the Humanitarian Turn 4.1 The long stalemate The very first resolution of the United Nations General Assembly, adopted in January 1946, called for the elimination of atomic weapons from national armaments. Nothing of the kind followed. Instead, the world built a system that managed nuclear weapons rather than abolishing them. The Treaty on the Non-Proliferation of Nuclear Weapons, opened for signature in 1968, was the centrepiece of that system. It rested on a bargain in three parts. Non-nuclear states promised not to acquire the weapons. They received in return a guarantee of access to peaceful nuclear technology under safeguards. And the five states that had already tested devices before 1967 promised, in Article VI, to pursue negotiations in good faith on effective measures relating to nuclear disarmament. For many years the majority of the world's states waited for that third promise to be honoured. Arsenals fell sharply after the Cold War, but they did not disappear, and by the 2010s the trend had reversed. #modernisation programmes in every nuclear armed state pointed towards arsenals maintained indefinitely. According to the Stockholm International Peace Research Institute (2025), roughly twelve thousand warheads remained in existence at the start of 2025, of which about nine and a half thousand sat in military stockpiles available for use, and the number available for use has risen every year since 2017. The 2010 Review Conference of the #NPT produced a final document that spoke of the catastrophic humanitarian consequences of any use of nuclear weapons. That phrase, almost unnoticed at the time, turned out to be a hinge. It gave frustrated states a vocabulary that was not about strategic parity or throw weight, and that the nuclear powers had themselves signed. 4.2 The Humanitarian Initiative Between 2013 and 2014 three intergovernmental conferences examined what would actually happen if nuclear weapons were used. Norway hosted the first in Oslo in March 2013. Mexico hosted the second in Nayarit in February 2014. Austria hosted the third in Vienna in December 2014. The #scientific_evidence presented at those conferences was the decisive factor. Medical experts explained that no state and no international organisation has the capacity to respond adequately to a nuclear detonation in a populated area. Climate scientists presented models of #nuclear_winter through soot injection into the stratosphere and the resulting collapse in global agricultural output. Survivors of the bombings of Hiroshima and Nagasaki, the #hibakusha, and survivors of #nuclear_testing in the Pacific, in Kazakhstan, in Algeria and in Australia described what had happened to their bodies, their families and their land. The Nayarit conference reached a conclusion that became the strategic core of everything that followed: the prohibition of a category of weapon has historically preceded and stimulated its elimination, not the other way round. That reversed the standard sequencing assumption of arms control, which held that you must first build trust and verification capacity and only then outlaw the weapon. At the close of the Vienna conference, Austria issued what became known as the Humanitarian Pledge, committing to work to fill the legal gap for the prohibition and elimination of nuclear weapons. The #Humanitarian_Pledge was eventually endorsed by well over a hundred states. Kmentt (2021) describes this as the moment at which the initiative stopped being an awareness campaign and became a legislative project. 4.3 From pledge to mandate The 2015 NPT Review Conference failed to agree a final document. For the states behind the pledge, this confirmed that the consensus based forums of the existing regime were structurally incapable of delivering disarmament, because any single nuclear armed state could block progress. Egeland (2022, 2024) argues that this realisation, more than any change in the material environment, explains the shift in strategy. In 2016 the General Assembly convened an open-ended working group in Geneva. It recommended negotiations in 2017 on a legally binding instrument to prohibit nuclear weapons leading towards their total elimination. In October 2016 the First Committee adopted a resolution to that effect by 123 votes to 38 with 16 abstentions, and in December the General Assembly confirmed the mandate. Negotiations were scheduled for two sessions in New York: 27 to 31 March and 15 June to 7 July 2017. The #nuclear_armed_states boycotted. Most of their allies boycotted with them. The Netherlands, under pressure from its own parliament, attended and voted against the final text. Around 130 states took part. 5. Negotiation and Adoption 5.1 The conference The negotiating conference was presided over by Ambassador Elayne Whyte Gomez of Costa Rica. Two features of the process deserve attention because they explain the shape of the final text. First, #civil_society was not merely present but structurally embedded. The International Campaign to Abolish Nuclear Weapons coordinated dozens of organisations, supplied legal drafting expertise, tracked national positions and organised survivor testimony. The campaign received the Nobel Peace Prize in October 2017 for that work. Acheson (2021) documents how this coalition operated. In 2024 the Nobel Peace Prize went to Nihon Hidankyo, the Japanese organisation of atomic bomb survivors, a recognition of the same movement from a different angle. Second, the negotiation ran on a simple #majority_voting basis rather than consensus. This is the single most important procedural fact about the treaty. Consensus rules in the Conference on Disarmament had produced nothing for two decades because any one state could veto. By moving to the General Assembly and to majority voting, the sponsors removed the veto and made adoption possible. Critics regard this as the treaty's original sin. Supporters regard it as the only reason it exists. 5.2 The bargains inside the text Three internal compromises shaped the treaty. The first concerned peaceful nuclear energy. Many Global South states would not accept any text that could later be used to restrict their access to civil nuclear technology. Futter and Samuel (2024) show that the preamble's affirmation of the inalienable right to peaceful uses was the condition of their support. Without it the treaty would not have reached 122 votes. The second concerned the relationship with the NPT. Several delegations were determined that the new instrument should be presented as an implementation of Article VI of the NPT rather than as a rival to it. The preamble accordingly reaffirms the NPT as the cornerstone of the regime. The third concerned nuclear testing and its legacy. Kazakhstan, Kiribati and other states with test sites on their territory insisted that the treaty address past harm, not only future prohibition. This produced the #positive_obligations in Articles 6 and 7, which are discussed in Section 7 and which have no precedent in nuclear law. 5.3 The vote On 7 July 2017 the conference adopted the text: 122 in favour, one against, one abstention. The Netherlands voted against. Singapore abstained. The treaty opened for signature at the United Nations in New York on 20 September 2017. 6. The Legal Architecture of the Treaty The treaty has a preamble of twenty-four paragraphs and twenty articles. What follows is a plain language walk through the operative provisions, with attention to the points that matter most for legal analysis. 6.1 The preamble Preambles are not directly binding, but they guide interpretation and they carry political meaning. This one does several things at once. It states the catastrophic humanitarian consequences of any use of nuclear weapons. It records the ethical imperative of disarmament. It acknowledges the disproportionate impact of ionising radiation on women and girls. It acknowledges the disproportionate impact of nuclear testing on indigenous peoples. It names the hibakusha and the victims of testing. And it affirms compliance with existing law: the Charter, international humanitarian law, human rights law, the NPT, the test ban treaty and the regional zones. The reference to international humanitarian law is doing significant work. The treaty is not asserting a wholly new legal claim out of nothing. It is asserting that principles already accepted by all states, in particular the rules of distinction, proportionality and the prohibition on causing unnecessary suffering, cannot be reconciled with the use of a weapon whose effects cannot be contained in space or time. 6.2 Article 1: the prohibitions This is the heart of the treaty. Each state party undertakes never, under any circumstances, to: develop, test, produce, manufacture, otherwise acquire, possess or stockpile nuclear weapons or other nuclear explosive devices; transfer such weapons or control over them to any recipient; receive the transfer of or control over such weapons; use or threaten to use nuclear weapons; assist, encourage or induce anyone to engage in any prohibited activity; seek or receive any assistance from anyone to engage in a prohibited activity; allow the stationing, installation or deployment of nuclear weapons on its territory or at any place under its jurisdiction or control. Three elements are worth pausing on. The #threat_of_use. No previous nuclear treaty prohibited threatening to use the weapons. This provision goes to the core of #nuclear_deterrence, which is a policy of conditional threat. If threatening use is unlawful for a party, then a party cannot maintain a deterrent posture, and cannot coherently endorse someone else's. Krasno and Szeli (2024) treat this as the treaty's boldest legal move and also its most contested. Assistance, encouragement and inducement. This clause is what makes membership incompatible with participation in extended deterrence. A state that hosts foreign nuclear weapons, that provides delivery aircraft, that permits transit, or that formally endorses the use of nuclear weapons on its behalf in an alliance doctrine would run into difficulty. Hummel (2024) shows how sharply this bites in the Belgian debate. The stationing ban. A state that has another state's weapons on its territory must remove them if it joins, subject to the deadlines discussed below. 6.3 Article 2: declarations Each state party must submit a #declaration to the Secretary-General within thirty days of the treaty entering into force for it, stating whether it owned or possessed nuclear weapons and eliminated its programme before entry into force, whether it currently owns or possesses them, and whether it hosts another state's weapons on its territory. This creates a baseline record. 6.4 Article 3: safeguards States parties that have no nuclear weapons must, at a minimum, maintain the #IAEA safeguards obligations they already have. A party that has not yet concluded a comprehensive safeguards agreement must do so within a specified period after entry into force for that state. This provision matters because a common criticism of the treaty is that it weakens the verification standards of the non-proliferation regime. In fact it sets a floor and requires the laggards to catch up, although critics reply that the floor should have been set higher by requiring the Additional Protocol of all parties. 6.5 Article 4: elimination This is the article that would matter most if a nuclear armed state ever joined, and it is the article most often mischaracterised. It sets out two pathways. Under the first pathway, a state that eliminated its nuclear weapons programme before entry into force for it must cooperate with a competent international authority designated to verify the irreversible elimination. Under the second pathway, a state that still owns or possesses nuclear weapons when it joins must immediately remove them from operational status and destroy them as soon as possible, according to a legally binding, time-bound plan negotiated with a competent international authority and submitted to the states parties for approval. Under a third route, a state that hosts another state's weapons must ensure their prompt removal. Two points follow. First, the treaty does not name the competent international authority. This is deliberate: the parties did not wish to prejudge an institutional design that would have to be negotiated with a disarming state anyway. Critics call this a hole. Supporters call it flexibility. Second, the first Meeting of States Parties in 2022 filled in the deadlines that Article 4 left to be determined. It agreed a ten year period for the destruction of a joining state's own arsenal, with the possibility of extension, and a ninety day period for the removal of another state's weapons stationed on a party's territory. That decision converted an abstract framework into something with a schedule. The honest assessment is that #verification under Article 4 remains the treaty's weakest technical area, and the states parties know it. The Scientific Advisory Group established in 2023 exists partly to address this. 6.6 Article 5: national implementation Each party must adopt the #national_implementation measures necessary to implement its obligations, including legislation to prevent and suppress prohibited activity by persons or on territory under its jurisdiction or control. In practice this means criminal law. Very few parties have yet passed comprehensive implementing legislation, which is one of the clearest measurable gaps in the treaty's first five years. 6.7 Articles 8 to 20: institutional and final clauses Meetings of States Parties are to be convened biennially, and Review Conferences every six years. Amendments require a two-thirds majority. Disputes are to be settled by peaceful means. The treaty is open to all states. It permits no #reservations, which is unusual and significant: a state cannot join while carving out an exception for itself. #Withdrawal requires twelve months notice and a statement of the extraordinary events that justify it, and, importantly, a withdrawal does not take effect while the withdrawing state is a party to an armed conflict. The treaty is of unlimited duration. The prohibition on reservations deserves emphasis. It means that #states_parties cannot join the treaty and simultaneously preserve an alliance commitment to nuclear use. Membership is all or nothing. This is a strength in terms of normative clarity and a weakness in terms of #universalization, because it removes the diplomatic device by which reluctant states are usually brought inside a regime. 7. Positive Obligations: Victim Assistance and Environmental Remediation If Article 1 is the treaty's most famous provision, Articles 6 and 7 are its most original. 7.1 What the articles require Article 6 obliges each state party, in respect of individuals under its jurisdiction who are affected by the use or testing of nuclear weapons, to provide adequate assistance in accordance with international humanitarian law and human rights law. That assistance must be age and gender sensitive, must be provided without discrimination, and must include medical care, rehabilitation and psychological support, together with provision for social and economic inclusion. Article 6 further requires each party to take necessary and appropriate measures towards the environmental remediation of contaminated areas under its jurisdiction or control. Article 7 obliges parties to cooperate with one another and creates a duty of #international_cooperation and creates a duty of international cooperation and assistance. It also states that a party that has used or tested nuclear weapons has a responsibility to provide adequate assistance to affected states parties, for the purpose of victim assistance and environmental remediation. That responsibility is framed carefully. It does not create liability in the sense of a compensation claim enforceable in a court, and it does not bind non-parties. But it does put on the record, in a treaty, the proposition that those who caused the harm owe something to those who suffered it. 7.2 Why this matters Nuclear harm is not a hypothetical. More than two thousand #nuclear_explosions were conducted between 1945 and 1996. Communities in Kazakhstan, in the Marshall Islands, in French Polynesia, in Algeria, in Australia and in the American southwest live with the consequences: elevated cancer rates, birth defects, contaminated land and displacement that in some cases has never been reversed. Until 2017 no nuclear treaty said anything about these people. The design of Articles 6 and 7 draws on the model developed in the landmine and cluster munitions conventions, the reference model for #humanitarian_disarmament, where obligations to assist survivors and clear contaminated land are standard. Transplanting that model into the nuclear field is what commentators mean when they speak of the treaty as an instrument of humanitarian disarmament rather than of arms control. Arms control asks how many weapons a state may have. Humanitarian disarmament asks what the weapons do to people and who is responsible for the damage. 7.3 Implementation so far Progress has been real but slow. An #informal_working_group on Articles 6 and 7 was established at the first Meeting of States Parties and is co-chaired by Kazakhstan and Kiribati, two states with nuclear test sites on their territory. The second Meeting adopted reporting guidelines on a provisional basis and began work towards an international trust fund. The third Meeting in March 2025 advanced that work further and kept the trust fund on the agenda for the Review Conference. The obstacles are practical. Affected states are frequently poor. Remediation of a contaminated test site is extraordinarily expensive, and in some cases the technology to do it safely does not exist. The states that conducted the testing are not parties to the treaty and are under no treaty duty to pay. This is where the gap between the treaty's moral ambition and its material capacity is widest, and it is the area in which honest scholarship is most needed and least available. The concept of #nuclear_justice, which frames these obligations as a matter of redress for historic wrongs rather than charity, is gaining ground in the literature and in the diplomacy alike. 8. The Institutional Life of the Treaty since 2021 A treaty is not only a text. It is a set of meetings, working groups, reports and quiet arguments. The institutional record since entry into force tells us a good deal about what kind of instrument this is becoming. 8.1 The first Meeting of States Parties, Vienna, June 2022 The first #Meeting_of_States_Parties took place in Vienna from 21 to 23 June 2022, under the presidency of Alexander Kmentt of Austria. Its timing was extraordinary because of the war in #Ukraine. Russia had invaded Ukraine four months earlier and had made explicit reference to its nuclear forces. The meeting adopted a political declaration and the #Vienna_Action_Plan, a fifty point programme covering universalisation, victim assistance, remediation, verification and the treaty's relationship with other instruments. The declaration condemned any and all nuclear threats, whether explicit or implicit, and stated that any use or threat of use of nuclear weapons violates international law including the Charter. Arndt, Horovitz and Onderco (2023) argue that the Russian attempt at nuclear coercion in 2022 largely failed to achieve its coercive aims, and one plausible reason is precisely that a large group of states had built a normative vocabulary with which to reject such threats publicly and immediately. The meeting also took the decisions on Article 4 deadlines described above, and established the informal working groups that now do most of the treaty's real work between sessions. Several observer states attended, including a number of NATO members and other allies of nuclear armed states. Their presence, while carefully framed as observation rather than endorsement, mattered. 8.2 The second Meeting of States Parties, New York, November and December 2023 The second meeting, presided over by Mexico, was held at United Nations Headquarters from 27 November to 1 December 2023. It adopted a declaration on upholding the prohibition and averting catastrophic consequences, and it took three notable decisions. First, it launched a #consultative_process on the security concerns of states. This was a direct response to the most serious criticism of the treaty, namely that it ignores the security problems that lead states to want nuclear weapons in the first place. The parties decided to engage that criticism rather than dismiss it, and to develop arguments about how security can be maintained without nuclear deterrence. Second, it moved forward on the international trust fund for victim assistance and environmental remediation. Third, it consolidated the #Scientific_Advisory_Group, a standing body of independent experts mandated to advise the parties on scientific and technical questions, including verification, humanitarian consequences and risk. This was an attempt to answer the charge that the treaty is a moral gesture unsupported by technical seriousness. 8.3 The third Meeting of States Parties, New York, March 2025 The third meeting took place from 3 to 7 March 2025. Its declaration was titled around strengthening commitment to a world free of nuclear weapons amid rising global instability, a title that captures the mood accurately. The meeting expanded the mandate of the working group on universalisation, advanced the trust fund, initiated a review of the Scientific Advisory Group's terms of reference, and elected #South_Africa to preside over the first #Review_Conference. Onderco and Vignoli, writing in 2025 on the basis of their statement analysis, noted that four years after entry into force the parties remain divided about what the treaty fundamentally is. For some it is primarily a humanitarian and emancipatory project. For others it is a security instrument that must eventually engage the nuclear armed states on their own terms. That division is not fatal, but it is real, and it will shape the Review Conference. 8.4 The first Review Conference, 2026 The first Review Conference is scheduled for 30 November to 4 December 2026 at United Nations Headquarters, with South Africa in the chair. Review Conferences under Article 8 are meant to assess the operation of the treaty and progress towards its purpose. This will be the moment at which the parties must show what five years of implementation have produced: how many states have passed implementing legislation, whether the trust fund exists in more than name, whether the verification architecture has advanced beyond principle, and whether the membership has continued to grow. 9. Membership: Who Has Joined and Who Has Not 9.1 The numbers As of mid-2026, seventy-five states have ratified or acceded to the treaty, with a further group of signatories that have not yet completed ratification. Counting parties and signatories together, the total passed the halfway mark of the world's states during 2025. A further group of states, identified by observers on the basis of their votes on the annual General Assembly resolution supporting the treaty, are supportive without having joined, bringing overall supportive states to around seventy percent of the world's total. 9.2 The geography The pattern is strikingly regional, and it is not random. The #Americas and #Africa lead. Almost every state in Latin America and the Caribbean is a party or a supporter, and support across Africa is very high. Both regions already sat inside nuclear weapon free zones, under the Treaty of Tlatelolco and the Treaty of Pelindaba respectively. Onderco and Vignoli (2024) provide statistical support for what this suggests: membership of a regional zone socialises states into a particular way of thinking about nuclear weapons, and that socialisation carries over into positions on the ban. Oceania and Southeast Asia show strong but uneven support. Central Asia is supportive, with Kazakhstan playing a leading role because of its testing legacy. #Europe is the outlier. Around two thirds of European states oppose the treaty, and the great majority of the world's #umbrella_states are European. This is the single clearest structural obstacle to universalisation. It is not an accident of culture. It is a consequence of alliance architecture: NATO's strategic concept describes the alliance as a nuclear alliance, and several European states host American weapons on their territory. 9.3 The absent No nuclear armed state has joined. None is likely to in the foreseeable future. No state that participates in extended nuclear deterrence has joined either. The case of #Japan is the most poignant and the most instructive. Nishida (2024) explains it as the product of a balancing act between two vectors that pull in opposite directions: a humanitarian imperative rooted in the experience of 1945 and sustained by strong public anti-nuclear sentiment, and a security imperative rooted in a deteriorating regional environment and reliance on the American nuclear guarantee. Japan has praised the treaty at the highest political level while declining to attend even as an observer. Mukai (2025) argues that this position is becoming untenable and that Japan must move from vagueness to concreteness. Kanno and Miwa (2024) find that Japanese public support for the treaty is real but responsive to perceived threat, which suggests that the domestic politics of joining will become harder, not easier, if the regional situation worsens. 9.4 What membership costs Onderco and Vignoli (2025) find that states join treaties more slowly when the obligations are heavy and when they are engaged in rivalries. The ban treaty imposes heavy obligations and permits no reservations. It should therefore be expected to grow steadily rather than dramatically, drawing first on states with low security exposure and existing regional commitments, then more slowly on states with more complicated positions. That is precisely the pattern the data show. Understanding this prevents both premature triumphalism and premature despair. 10. The Treaty and the Non-Proliferation Regime 10.1 The compatibility question The most persistent legal criticism of the treaty is that it undermines the #non_proliferation_regime. The argument runs as follows: the NPT is the cornerstone of the regime; the ban treaty creates a competing legal framework with weaker verification standards and a rival diplomatic forum; it splits the non-nuclear weapon states into two camps; and it therefore weakens the very instrument on which non-proliferation depends. The response of the treaty's supporters is that there is no legal conflict at all. Nothing in the ban treaty permits what the NPT forbids or forbids what the NPT requires. The prohibition treaty imposes additional obligations on states that voluntarily accept them, which is entirely ordinary in treaty law. Its preamble reaffirms the NPT as the cornerstone. It requires safeguards as a floor. And it presents itself as an effective measure in fulfilment of NPT Article VI, which is exactly what Article VI calls for. On the legal merits, the supporters have the better argument. There is no incompatibility between the two texts. The real dispute is political rather than legal, and it concerns forum and framing rather than obligation. 10.2 The deeper problem Tannenwald (2024) offers the sharpest diagnosis. The non-proliferation regime, she argues, is being undermined from within by the failure of its own promise. Non-nuclear states accepted permanent non-possession in 1968 in exchange for a commitment to disarmament. Fifty-eight years later the arsenals are being expanded rather than eliminated, arms control agreements between the largest possessors have lapsed or been suspended, and the review process repeatedly fails to reach agreement. In that context, the ban treaty is not the cause of division. It is a symptom of it. Gibbons (2022) shows how much the coherence of the regime has historically depended on active American diplomatic effort, and Gibbons (2024) describes the American campaign against the ban treaty as a case of a great power resisting an emerging norm rather than as a neutral defence of the existing order. That campaign has included direct lobbying of states considering ratification. It is a measure of how seriously the nuclear armed states take a treaty they publicly dismiss as irrelevant. 10.3 The realistic relationship The most sensible way to view the two instruments is as different tools for different tasks. The NPT manages proliferation and preserves the safeguards system. It contains the nuclear armed states, at least formally. The ban treaty carries the disarmament obligation that the NPT states but never operationalised, and it does so in a forum where the nuclear armed states cannot exercise a veto. Neither can substitute for the other. A world with only the ban treaty would have no proliferation management. A world with only the NPT has, on the evidence of five decades, no disarmament. 11. The Case Against the Treaty Serious scholarship requires stating the opposing case in its strongest form. The arguments against the treaty are not frivolous and students should be able to reproduce them accurately. 11.1 It disarms no one The most obvious #criticism is also the most powerful. Not a single warhead has been dismantled because of this treaty, and the number available for use has risen every year since it was negotiated. A prohibition that binds only those who never had the weapons in the first place can be characterised as a form of moral self-congratulation. Krasno and Szeli (2024) acknowledge this frankly even while defending the instrument. 11.2 It ignores security #deterrence advocates argue that nuclear weapons exist because states face existential threats from other states, and that these threats do not disappear because a treaty says the weapons are prohibited. On this view the treaty addresses the symptom and not the cause. A state that renounces its nuclear guarantee without any change in its threat environment has not become more moral; it has become more vulnerable, and possibly more likely to be attacked. The parties themselves took this criticism seriously enough to establish a consultative process on security concerns in 2023. 11.3 Its verification is thin Article 4 does not name the competent international authority and does not specify a verification methodology. Critics argue that a disarmament treaty without a credible verification regime is not a disarmament treaty at all. Supporters respond that verifying the dismantlement of an actual arsenal would require negotiation with the disarming state in any case, and that designing the machinery in advance would have been an academic exercise. Both positions have force. The parties are attempting to close the gap through the Scientific Advisory Group, but the work is at an early stage. 11.4 It fractures the regime The claim that the treaty divides non-nuclear weapon states and complicates the NPT review process is empirically defensible. Review conferences since 2017 have indeed featured sharp disagreement over the ban treaty. Whether this division is caused by the treaty or merely revealed by it is the crux of the argument, and it cannot be settled by assertion. 11.5 A response A fair assessment would concede the first criticism, engage seriously with the second and third, and reject the fourth. The treaty does not disarm anyone. It was never going to. Kmentt (2021) and Acheson (2021) both make clear that the aim was to change the legal and moral status of the weapons in the expectation that changed status would eventually change behaviour, on the model of the landmine and cluster munitions bans, where states that never joined nonetheless stopped using the weapons because using them became politically intolerable. Whether that model transfers to nuclear weapons, which are held for reasons of existential security rather than battlefield convenience, is the open question of the field. Egeland (2022) is right to warn that the analogy may not hold. 12. Normative and Practical Impact 12.1 Stigmatisation and its limits The #theory_of_change behind the treaty is #stigmatisation. Make the weapon illegal for a majority of states, attach the language of atrocity to it, and over time make possession politically costly, financially awkward and diplomatically embarrassing. There is evidence that this is working at the margins. Financial institutions in several countries have pursued #divestment from companies involved in nuclear weapons production, citing the treaty in their policies. Hundreds of #cities have endorsed calls for their governments to join. Parliamentary motions in umbrella states have multiplied. #public_opinion in several #NATO members shows majority support for joining, even where governments refuse. There is also evidence that it is not working where it counts. Arsenals are growing. France has announced an increase. China, India, Pakistan, North Korea and Russia are all expanding. Discussions about acquiring or hosting nuclear weapons have intensified in several non-nuclear states in Europe and East Asia. Egeland (2024) is right that scholars who focus only on delegitimisation miss the parallel and well funded work of relegitimisation. The nuclear order is not passively awaiting stigmatisation. It is fighting back, and at present it is winning on the material measures. 12.2 Discursive effects Where the treaty has clearly succeeded is in changing the terms of debate. Before 2013 the humanitarian consequences of nuclear weapons were largely absent from official disarmament discourse, which spoke the language of stability, parity and step-by-step process. That is no longer true. Even states that reject the treaty now feel obliged to argue in humanitarian terms, and the Group of Twenty statement of 2023 that the use or threat of use of nuclear weapons is inadmissible would have been unimaginable a decade earlier. This is not nothing. #norms move through language before they move through behaviour. 12.3 The customary law question Some advocates argue that the treaty is building towards a rule of #customary_international_law binding on all states, including non-parties. This claim should be handled with care. Customary law requires widespread and representative state practice together with a belief that the practice is legally required. The states that matter most for this particular rule, namely those that possess the weapons, are persistently objecting, loudly and in writing. #persistent_objection is a recognised barrier to the formation of a customary rule against the objecting state. The honest position is that the treaty contributes to a normative trend without yet creating custom, and that responsible scholarship should not overstate the point. Overstating it damages the credibility of the argument. 12.4 The effect on nuclear threats The clearest practical effect may be the one that is hardest to measure. When Russia signalled nuclear readiness in 2022, the response was immediate, coordinated and framed in the vocabulary the humanitarian initiative had built. Bollfrass and Herzog (2022) and Arndt, Horovitz and Onderco (2023) both suggest that the coercive value of nuclear threats has been lower than the threatening state expected. It is at least plausible that a world in which 122 states had recently declared such threats unlawful is a world in which such threats work less well. It is also possible that other factors explain the outcome entirely. The honest verdict is: suggestive, unproven. 13. Discussion: Obstacles and Pathways 13.1 The structural obstacle The central obstacle is not legal and not technical. It is that nuclear weapons are embedded in the security architecture of the most powerful states and of thirty-three of their allies. Those states will not disarm because a treaty asks them to. They will disarm, if ever, when they conclude that the weapons no longer serve their interests, or when domestic politics makes retention untenable. The treaty's theory is that it can contribute to bringing about that second condition. This is a long term theory, and it should be evaluated on a long timescale. The chemical weapons ban took decades. The abolition of slavery, the analogy Egeland (2024) treats with appropriate scepticism, took more than a century. Judging the ban treaty a failure five years after entry into force is like judging a tree a failure five years after planting. 13.2 The internal obstacle The parties are not united about what they are doing. Onderco and Vignoli (2024) demonstrate a genuine cleavage between those who see the project as humanitarian and emancipatory and those who see it as an eventual security instrument. This matters for strategy. A humanitarian framing prioritises victim assistance, remediation, survivor testimony and civil society mobilisation. A security framing prioritises engaging umbrella states, developing verification, and building arguments about how a state can be safe without nuclear weapons. Both are necessary. Doing both well requires resources that most of the parties, which are largely small and not wealthy, do not have. #financing is a real and underexamined constraint. The treaty's costs fall on its parties, and the majority of them have limited budgets. A treaty regime that cannot afford to run its own meetings and working groups properly will struggle to build the technical credibility it needs. 13.3 Pathways forward Four pathways appear plausible on the current evidence. Deepen #implementation. The most credible thing the parties can do in the next five years is to demonstrate that the treaty produces action. That means national implementing legislation under Article 5, functioning reporting under Articles 6 and 7, and a trust fund that actually distributes money to affected communities. A treaty that visibly helps people is far harder to dismiss than one that only issues declarations. Build the verification architecture. The Scientific Advisory Group should produce a serious, technically literate model for how Article 4 would work if a nuclear armed state ever chose to disarm. Even if that day is distant, having the answer ready removes one of the strongest arguments against the treaty and signals seriousness. Engage rather than shame the umbrella states. Nishida (2024) makes the case for #strategic_empathy: a genuine willingness to listen to the security concerns of nuclear dependent states rather than dismissing them as bad faith. The consultative process on security concerns launched in 2023 is the right instrument. The prize is large. If even one significant umbrella state joined, the political character of the treaty would change overnight. Work through domestic politics. Hummel (2024) and Kanno and Miwa (2024) both show that the decisive arena is national. #parliaments, cities, financial institutions, medical associations and public opinion are where the ban movement has its comparative advantage, and where the nuclear armed states cannot simply veto. 13.4 The scenario that would change everything None of the above is likely to move a nuclear armed state in the near term. The scenario that would change everything is the one nobody wants: an actual detonation. The humanitarian initiative was built on the argument that the consequences of use are unacceptable. If use ever occurred, the argument would be made for it in the worst possible way. It is worth being clear-eyed about this. The treaty exists precisely so that the world does not have to learn the lesson that way. 14. Limitations of This Study This article is a synthesis, not primary research. It relies on published scholarship and on official documents, and it does not include interviews, archival work or original data collection. Three specific limitations should be flagged. First, the evidence on the treaty's causal effects is thin, and the article has tried to say so rather than to fill the gap with confident assertion. Claims about stigmatisation, about the failure of nuclear coercion, and about the emergence of customary norms are all plausible and all underdetermined by the available evidence. Second, the literature is not neutral. Egeland and Pelopidas (2025) document the extent to which nuclear policy analysis is funded by actors with a material interest in the perpetuation of nuclear deterrence. That finding cuts both ways: it should make readers cautious about work that dismisses the treaty, and it should also make them cautious about advocacy scholarship that overstates its achievements. This article has tried to read both sides critically. Third, the situation is moving. The first Review Conference in late 2026 will produce decisions that this article cannot anticipate. Readers should treat the empirical claims here as accurate to mid-2026 and should verify current figures before relying on them. 15. Conclusion and Recommendations 15.1 Conclusion The Treaty on the Prohibition of Nuclear Weapons did something that the international system had failed to do for seventy years. It made the possession of nuclear weapons unlawful in a treaty of general application, for those willing to accept the obligation. It did so through an unusual coalition of small and middle powers, Global South states with their own developmental interests to protect, survivors of nuclear violence, and an organised civil society campaign, working in a forum where the great powers could not exercise a veto. The treaty's achievements are real but specific. It closed a gap in the #law. It gave a name and a legal status to a category of harm that the nuclear order had ignored. It created the first international obligations towards the victims of nuclear use and testing. It made the threat of use unlawful for its parties, which puts the doctrine of nuclear deterrence itself in legal question in a way that no earlier instrument had done. And it changed the vocabulary of the debate so thoroughly that even its opponents now argue on its terms. Its limits are equally real. It has disarmed no one. Its membership excludes every state that possesses the weapons and every state that shelters under them. Its verification provisions are incomplete. Its financing is precarious. Arsenals are growing, and the strategic environment is deteriorating rather than improving. The correct conclusion is therefore neither celebration nor dismissal. The treaty is best understood as a long term normative project whose success or failure cannot be assessed on a five year horizon. It has completed the emergence phase of the norm it carries. Whether that norm cascades depends on decisions that will be taken in national capitals, in parliaments, and in the strategic calculations of states that are not yet listening. The first Review Conference in 2026 will not settle the question. But it will show whether the parties have built an institution capable of surviving long enough for the question to be settled. 15.2 Recommendations For states parties. Prioritise visible implementation over further declaration. Pass Article 5 legislation. Fund and launch the international trust fund. Publish reporting under Articles 6 and 7. Give the Scientific Advisory Group a mandate and a budget adequate to produce credible verification work. For signatories that have not ratified. Complete ratification. The gap between signature and ratification is now a significant share of the treaty's potential membership and is a soft target for opponents. For umbrella states. Attend as observers. Observation costs nothing legally and signals that the security concerns of allied states deserve to be heard inside the process rather than shouted at from outside it. For researchers. The field needs independent, non-advocacy work on three questions: what national implementation actually looks like on the ground; whether victim assistance obligations are producing measurable benefit for affected communities; and how far, if at all, the treaty has altered the calculations of non-parties. Doctoral students looking for a genuinely open question could do worse than start there. For students. Read the #treaty_text itself. It is twenty articles long and written in accessible language. Almost everything written about it, including this article, is a commentary on a text that takes under an hour to read for yourself. Hashtags #TPNW #nuclear_disarmament #nuclear_weapons_ban #humanitarian_disarmament #international_law #NPT #nuclear_deterrence #ban_the_bomb #hibakusha #nuclear_justice #global_nuclear_order #arms_control #ICAN #treaty_on_the_prohibition_of_nuclear_weapons #peace_studies References Acheson, R. (2021). Banning the Bomb, Smashing the Patriarchy. Lanham, MD: Rowman and Littlefield. Arndt, A., Horovitz, L., and Onderco, M. (2023). Russia's failed nuclear coercion against Ukraine. The Washington Quarterly, 46(3), 167-184. Bollfrass, A. K., and Herzog, S. (2022). The war in Ukraine and global nuclear order. Survival, 64(4), 7-32. https://doi.org/10.1080/00396338.2022.2103255 Egeland, K. (2021). Nuclear weapons and adversarial politics: Bursting the abolitionist consensus. Journal for Peace and Nuclear Disarmament, 4(1), 107-115. Egeland, K. (2022). A theory of nuclear disarmament: Cases, analogies, and the role of the non-proliferation regime. Contemporary Security Policy, 43(1). Egeland, K. (2024). The Struggle for Abolition: Power and Legitimacy in Multilateral Nuclear Disarmament Diplomacy. Abingdon: Routledge. Egeland, K., and Pelopidas, B. (2025). No such thing as a free donation? Research funding and conflicts of interest in nuclear weapons policy analysis. International Relations. https://doi.org/10.1177/00471178221140000 Futter, A., and Samuel, O. (2024). Accommodating Nutopia: The nuclear ban treaty and the developmental interests of Global South countries. Review of International Studies, 50(5), 799-820. https://doi.org/10.1017/S0260210523000396 Gibbons, R. D. (2022). The Hegemon's Tool Kit: US Leadership and the Politics of the Nuclear Nonproliferation Regime. Ithaca, NY: Cornell University Press. https://doi.org/10.1515/9781501764868 Gibbons, R. D. (2024). How great powers resist emerging norms: The United States and the Treaty on the Prohibition of Nuclear Weapons. Peace Review, 36(2), 335-344. Herzog, S. (2025). Why states join multilateral nuclear arms control, nonproliferation, and disarmament treaties. In Nuclear Politics and International Cooperation (pp. 33-49). Cham: Springer. Hummel, H. (2024). Contested nuclear sharing in Belgium: Domestic political dynamic for the nuclear weapons ban. Journal for Peace and Nuclear Disarmament. https://doi.org/10.1080/25751654.2024.2385171 Jadoon, U. I. (2021). The security impact of the Treaty on the Prohibition of Nuclear Weapons. In Nuclear Non-Proliferation in International Law, Volume VI: Nuclear Disarmament and Security at Risk (pp. 365-383). The Hague: T.M.C. Asser Press. Kanno, T., and Miwa, H. (2024). Military threats, international pressure, and public opinion on the nuclear ban treaty in Japan. Journal for Peace and Nuclear Disarmament, 7(2), 446-467. Kmentt, A. (2021). The Treaty Prohibiting Nuclear Weapons: How it was Achieved and Why it Matters. Abingdon: Routledge. Krasno, J., and Szeli, E. (2024). The Treaty on the Prohibition of Nuclear Weapons: Strengths, weaknesses and current status. Journal for Peace and Nuclear Disarmament, 7(2), 277-285. https://doi.org/10.1080/25751654.2024.2364967 Mukai, W. (2025). Japan and its complex position in the nuclear age: Moving from vagueness to concreteness. Journal for Peace and Nuclear Disarmament, 8(1), 36-56. https://doi.org/10.1080/25751654.2025.2518764 Nishida, M. (2023). Changing security environment in East Asia and its implications on Japan's nuclear policy. Journal for Peace and Nuclear Disarmament, 6(2), 327-345. Nishida, M. (2024). Logic (or illogic?) of Japan's aversion to TPNW and charting a path forward. Journal for Peace and Nuclear Disarmament, 7(2), 427-445. https://doi.org/10.1080/25751654.2024.2379085 Onderco, M. (2021). Networked Nonproliferation: Making the NPT Permanent. Stanford, CA: Stanford University Press. https://doi.org/10.11126/stanford/9781503628922.001.0001 Onderco, M., and Vignoli, V. (2024). Regional socialization and disarmament preferences: Explaining state positions on the nuclear ban treaty. Contemporary Security Policy, 45(3), 525-554. https://doi.org/10.1080/13523260.2024.2376416 Onderco, M., and Vignoli, V. (2025). Treaty legalization, security interests, and ratification of multilateral disarmament treaties. Conflict Management and Peace Science, 42(3), 223-244. Portela, C., Onderco, M., and Vignoli, V. (2025). Does CFSP co-ordination foster convergence? Voting behavior on nuclear weapons at the UN General Assembly. Contemporary Security Policy, 46(4), 1171-1196. Rajagukguk, M. H., and Siagian, R. C. (2024). Importance and challenges in TPNW ratification: Regional analysis of Oceania, the Americas, Europe, Asia, and Africa. Security Intelligence Terrorism Journal, 1(2), 80-91. https://doi.org/10.70710/sitj.v1i2.9 Smetana, M., and Onderco, M. (2025). Hope the Russians love their children too: Russian public support for the use of nuclear weapons after the invasion of Ukraine. Journal of Global Security Studies, 10(3), 1-20. Soeparna, I. I. (2023). Prohibition of nuclear weapons: Towards sustainable human security. Journal of Human Security, 19(1), 32. Stockholm International Peace Research Institute. (2025). SIPRI Yearbook 2025: Armaments, Disarmament and International Security. Oxford: Oxford University Press. Tannenwald, N. (2024). The nuclear nonproliferation regime as a failed promise: Contestation and self-undermining dynamics in a liberal order. Global Studies Quarterly, 4(2). Vignoli, V., and Onderco, M. (2025). Leader ideology and state commitment to multilateral treaties. The Review of International Organizations, 20(4), 1023-1052.

The Comprehensive Nuclear-Test-Ban Treaty (CTBT), opened for signature in September 1996, prohibits every nuclear weapon test explosion and every other nuclear explosion, whether the stated purpose is military or civilian. Thirty years later, the treaty occupies an unusual position in international law. It has almost universal political support, a large and technically advanced verification system, and a permanent secretariat in Vienna, yet it has never entered into force. This article offers a structured review of the treaty for students and early career researchers. It examines the historical road from the first atmospheric tests to the Geneva negotiations, the legal design of the treaty text, the institutional shape of the Preparatory Commission, and the four sensor technologies of the International Monitoring System. It then reviews recent peer reviewed work on seismic discrimination, machine learning, radionuclide background estimation, and on-site inspection readiness, and connects that technical literature to the political record: the United States Senate vote of 1999, the continued abstention of eight other Annex 2 states, the Russian withdrawal of ratification in 2023, and the statements made in late 2025 about a possible resumption of American testing. The article argues that the CTBT is best understood as a treaty whose normative and technical achievements have outrun its legal status, and that this gap is the central research problem for the coming decade. Three findings are highlighted. First, the verification regime already works at a level of sensitivity that the negotiators of the 1990s did not expect. Second, the absence of entry into force blocks the treaty's strongest compliance tool, the short notice on-site inspection, and leaves the ban resting on a moratorium that any single state can break. Third, the norm against testing is strong but not self-sustaining, and its survival now depends on political choices rather than on technical capability. Keywords: nuclear test ban; arms control; verification; International Monitoring System; on-site inspection; non-proliferation; disarmament; treaty law; nuclear norms; seismology 1. Introduction Between 1945 and 1996, states detonated more than two thousand nuclear devices. Most were underground, but hundreds were exploded in the atmosphere, under water, and even in space. Each explosion was a step in the refinement of a weapon, and each also left a physical trace in soil, air, bone, and water. The #comprehensive_nuclear_test_ban_treaty, adopted by the United Nations General Assembly in September 1996, was written to close that chapter. Its central promise is short and absolute: no nuclear weapon test explosion, and no other nuclear explosion, anywhere, by anyone, for any purpose. That last phrase matters more than students often realise. The treaty does not only ban military testing. It also bans so-called peaceful nuclear explosions, the civil engineering projects that both the United States and the Soviet Union once pursued with enthusiasm, digging canals and stimulating gas fields with nuclear devices. The negotiators understood that a bomb used to move earth is still a bomb, and that data from such a shot can improve a warhead just as easily as data from a weapons trial. By covering civilian purposes as well as military ones, the treaty removed a loophole that had weakened earlier agreements. The treaty is therefore ambitious in scope. It is also, in a formal legal sense, incomplete. Article XIV of the CTBT requires ratification by forty four named states before the treaty can enter into force. These are the states that took part in the negotiations and that possessed nuclear power or research reactors at the time. As of late 2025, one hundred and eighty seven states had signed and one hundred and seventy eight had ratified, but nine of the forty four listed states had still not deposited an instrument of ratification. Six of them had signed without ratifying, and three had not signed at all. The treaty therefore lives in a strange condition that some scholars call legal limbo and others call #provisional_application_by_practice. This article has four aims. The first is descriptive: to set out clearly what the treaty says, how it was built, and what its institutions do. The second is technical: to explain, in plain language, how the world actually detects a nuclear explosion, and how much the science has improved since 1996. The third is political: to explain why nine states have held back, and why the situation has grown more fragile since 2023. The fourth is analytical: to ask what the CTBT teaches us about the life of international norms when law and practice come apart. The article is written for students. It assumes no prior training in seismology or in treaty law. Technical terms are explained when they first appear. Where the literature disagrees, the disagreement is presented rather than smoothed over. Where the evidence is thin, that is said plainly. 2. Literature Review Research on the CTBT falls into three fairly separate communities, and one of the difficulties for students is that these communities rarely read each other. The first is the #geophysics community. It publishes in journals such as Pure and Applied Geophysics, the Journal of Seismology, and the Bulletin of the Seismological Society of America. Its questions are practical and technical: how small an explosion can be detected, how to distinguish a quarry blast from a bomb, how atmospheric conditions bend infrasound waves, how radioactive xenon released by hospitals and reactors creates a confusing background. Recent topical issues built around the CTBT Science and Technology conferences give a good map of this field (Mialle and Kalinowski 2025; Kalinowski and Mialle 2021; Kalinowski et al. 2023). The tone of this literature is confident. Detection thresholds keep falling, and the papers show a system that performs better each year. The second is the #arms_control_policy community, which publishes in the Journal for Peace and Nuclear Disarmament, the Nonproliferation Review, Survival, and Contemporary Security Policy. Its questions are political: why do states ratify, what does the treaty do for security, how does the test ban interact with the Nuclear Non-Proliferation Treaty and with the newer Treaty on the Prohibition of Nuclear Weapons. Work by Gibbons and Herzog (2022) on the strain placed on the non-proliferation regime by a more multipolar world, by Bollfrass and Herzog (2022) on the effect of the war in Ukraine on global nuclear order, and by Maitre (2025) on the prospects of the prohibition treaty, sets the frame within which the test ban must be read. Sato (2021) makes an argument that runs through this article: that the verification regime can build confidence between states even while the treaty remains unratified, and that the two questions should be analytically separated. The third community is #international_legal_scholarship, which asks what obligations bind a state that has signed but not ratified, what the Vienna Convention on the Law of Treaties says about withdrawing a ratification before entry into force, and whether a customary prohibition on testing has formed. This literature grew quickly after the Russian decision of 2023, because that decision raised a question with no clean precedent. There is a fourth strand that deserves mention, even if it is smaller: the #humanitarian_and_environmental literature on test victims and contaminated landscapes, produced largely by affected communities in Kazakhstan, the Marshall Islands, French Polynesia, Algeria, and the western United States. This work is often excluded from mainstream security journals, but it supplies the moral argument that made the ban politically possible. The gap this article addresses is the space between these communities. Technical papers usually assume that the political problem is someone else's business. Political papers usually treat verification as a solved black box. Neither assumption survives contact with the record. The claim of the 2019 and 2025 American compliance reports, that Russia and China may have conducted experiments above the zero yield threshold, is simultaneously a technical claim about detection limits and a political claim about intent. It cannot be evaluated inside one discipline. 3. Methodology and Scope This is a qualitative, document based review. It draws on three kinds of material. The first is the treaty text itself, together with its Protocol and two Annexes, and the resolution of 19 November 1996 that established the Preparatory Commission. These are primary legal sources and they are quoted only in paraphrase. The second is the recent peer reviewed technical literature, mainly from 2021 to 2025, on monitoring and verification. Papers were selected because they were published in indexed journals and because they addressed a distinct component of the verification regime: seismic location and discrimination, infrasound propagation, hydroacoustic coverage, radionuclide detection and background, on-site inspection technique, and machine learning applications. The third is the recent policy and legal literature on ratification, on the Russian withdrawal of ratification, and on the 2025 statements from Washington about resuming testing. Three limits should be stated honestly. First, some of the most important information about state behaviour is classified, so any claim about what happens at Novaya Zemlya, Lop Nur, or the Nevada National Security Site rests on open sources and satellite imagery, not on direct evidence. Second, the article does not attempt an original quantitative analysis of detection capability. Third, the political situation is moving quickly. Statements made in late 2025 have not yet produced a settled policy, and readers should check for developments after the date of writing. 4. Historical Background: From Trinity to Geneva 4.1 The age of open testing The first nuclear explosion, the Trinity test in New Mexico in July 1945, was followed within three weeks by the destruction of Hiroshima and Nagasaki. Testing then expanded rapidly. The United States tested in Nevada and in the Pacific. The Soviet Union tested in Kazakhstan and in the Arctic. The United Kingdom tested in Australia and later in Nevada. France tested in Algeria and then in French Polynesia. China tested at Lop Nur. By the mid 1950s the physical consequences were impossible to hide. Radioactive #fallout from atmospheric tests circled the globe. Strontium 90, chemically similar to calcium, appeared in milk and in children's teeth. The 1954 Castle Bravo test in the Marshall Islands produced a yield far larger than predicted, contaminated inhabited atolls, and irradiated the crew of a Japanese fishing boat. Public alarm grew into a genuine international movement, and that movement, not any strategic calculation, is what first put a test ban on the diplomatic agenda. 4.2 The partial ban and the threshold treaties The first result was the 1963 Partial Test Ban Treaty, which prohibited nuclear explosions in the atmosphere, under water, and in outer space, but permitted them underground. It was a #public_health victory rather than an arms control one. Testing did not stop. It simply moved below the surface, where it continued at a high rate for another three decades. The 1974 Threshold Test Ban Treaty limited underground explosions to one hundred and fifty kilotons, and the 1976 Peaceful Nuclear Explosions Treaty applied a similar cap to civil uses. Both agreements accepted the practice of testing and tried to regulate its scale. The idea of a total ban survived, but it was blocked by two arguments that would return again and again: that testing was necessary to keep an ageing arsenal reliable, and that a total ban could not be verified. 4.3 The window of the 1990s Two developments broke the deadlock. The first was political. The Cold War ended, the Soviet Union dissolved, and a moratorium spread: the Soviet Union stopped testing in 1990, the United States adopted a legislated moratorium in 1992, and China and France conducted their last tests in 1996. The second was scientific. #stockpile_stewardship, the combination of computer simulation, subcritical experiments, laboratory physics, and detailed surveillance of existing warheads, offered the nuclear weapon states a way to certify their arsenals without explosive testing. The technical objection weakened. Negotiations opened at the Conference on Disarmament in Geneva in 1994 and ran for more than two years. The chairman of the ad hoc committee, Ambassador Jaap Ramaker of the Netherlands, presented a final draft in June 1996. The Conference on Disarmament works by consensus, and consensus was not reached: India objected, above all to the entry into force clause, which it regarded as an unprecedented constraint on its sovereign right to decide whether to join a treaty. The text was therefore taken to the United Nations General Assembly, which adopted it in September 1996. The treaty opened for signature in New York on 24 September 1996 and collected seventy one signatures on the first day. The manner of its birth matters. A treaty that cannot pass by consensus in the negotiating forum, and that is then adopted in a different forum over the explicit objection of a state named in its own entry into force annex, begins life with a structural weakness built into it. Much of the treaty's later history follows from that first compromise. 5. The Legal Architecture of the Treaty 5.1 The basic obligation Article I contains the heart of the agreement. Each state party undertakes not to carry out any nuclear weapon test explosion or any other nuclear explosion, and to prohibit and prevent such explosions at any place under its jurisdiction or control. Each state party also undertakes not to cause, encourage, or in any way participate in such an explosion. The second limb is important. It closes the door on a state that might otherwise sponsor a test on someone else's territory. Two features of Article I deserve attention. The first is the absence of a #yield_threshold. The treaty does not say "no explosion above X kilotons". It says no explosion. The second is the absence of a definition of the word explosion. This is the single most consequential silence in the text, and section 10 of this article returns to it. 5.2 The organisation Article II establishes the Comprehensive Nuclear-Test-Ban Treaty Organization, with a Conference of States Parties, an Executive Council, and a Technical Secretariat, all seated in Vienna. On paper, the CTBTO comes into legal existence only when the treaty enters into force. In practice, a Preparatory Commission has existed since 1997, created by a separate resolution of the states signatories, and its Provisional Technical Secretariat performs almost all the functions that the treaty envisages for the permanent body. This is an unusual arrangement, and it is one of the reasons the treaty has achieved so much despite its status. 5.3 Verification Article IV sets out the #verification_regime. It has four components: the International Monitoring System, the International Data Centre, a process of consultation and clarification, and on-site inspections, together with voluntary confidence building measures. Article IV also obliges parties to cooperate in improving the regime and to examine the potential of additional technologies, such as satellite monitoring and electromagnetic pulse detection. Article V deals with measures to redress a violation and to ensure compliance. Article VI covers dispute settlement. Article VII covers amendment. Article VIII provides for a review conference ten years after entry into force. Article IX contains a withdrawal clause of the standard arms control type, allowing a party to leave if extraordinary events related to the subject matter have jeopardised its supreme interests. 5.4 Article XIV and the Annex 2 problem Article XIV is the clause that has defined the treaty's life. It provides that the treaty enters into force one hundred and eighty days after all forty four states listed in Annex 2 have deposited instruments of ratification. Annex 2 was constructed from a list, supplied by the International Atomic Energy Agency, of states that participated in the negotiations and that had nuclear power reactors or research reactors at the time. The intention was to guarantee that a test ban would bind everyone with the technical means to test. The effect has been to hand each of forty four states an absolute veto over the treaty's legal existence. Article XIV also provides that, if the treaty has not entered into force three years after opening for signature, a conference of ratifying states may be convened to examine what can be done to accelerate ratification. These #article_xiv_conferences have been held regularly since 1999, in Vienna and in New York, and the most recent took place in 2025. They produce declarations, appeals, and diplomatic pressure. They have not produced ratifications from the states that matter. The nine outstanding Annex 2 states, as of late 2025, are China, Egypt, India, Iran, Israel, North Korea, Pakistan, Russia, and the United States. Six of these have signed but not ratified; India, Pakistan, and North Korea have never signed. Russia belongs in this list only because it withdrew its ratification in November 2023, an event examined in section 11. 6. The Institutional Machinery 6.1 The Preparatory Commission The Preparatory Commission has a mandate to build the verification regime so that it is ready on the day the treaty enters into force. In practice it has done far more. It operates a global sensor network, processes the data, distributes it to national data centres, trains inspectors, and runs field exercises. Its Provisional Technical Secretariat employs several hundred staff in Vienna, and its budget is paid by the states signatories. The scale of what has been built is easy to underestimate. When complete, the International Monitoring System will consist of three hundred and thirty seven facilities: three hundred and twenty one stations and sixteen radionuclide laboratories. By October 2025, three hundred and seven of these had been certified, which is about ninety two percent of the planned network (Mialle and Kalinowski 2025). The stations sit in some of the least accessible places on Earth, from Antarctic ice to mid ocean islands, and they transmit data continuously through a dedicated satellite communications infrastructure. 6.2 National data centres Each state signatory may establish a #national_data_centre, which receives raw data and analysis products from Vienna and can perform its own independent assessment. This is a quiet but powerful design feature. It means that no single authority owns the interpretation of an event. When a suspicious signal appears, dozens of national laboratories examine the same data and can reach their own conclusions. Recent work from national centres in Kenya and Botswana shows how this capacity is used for domestic seismology as well as for treaty monitoring (Mulwa 2024; Simon et al. 2024). 6.3 Capacity building and the political function of science The Preparatory Commission also runs training programmes, workshops, and a biennial Science and Technology conference in Vienna that brings hundreds of researchers together. Sato (2021) argues that these activities function as #confidence_building_measures in their own right. States that share data, train together, and publish together find it harder to treat each other purely as adversaries. Whether this effect is strong enough to survive a sharp deterioration in great power relations is an open empirical question, and the years since 2022 are testing it. 7. The Verification Regime: Four Ways to Hear a Bomb The International Monitoring System is built on a simple insight. A nuclear explosion is a violent release of energy, and energy leaves the scene of an explosion in several forms. It shakes the ground. It pushes on water. It pushes on air. And, unless it is perfectly contained, it releases radioactive material. Four sensor networks were designed, one for each of these signatures. 7.1 Seismic monitoring The #seismic_network is the backbone. It consists of fifty primary stations, which send data continuously, and one hundred and twenty auxiliary stations, which can be queried when needed. Many of the primary stations are arrays: clusters of instruments spread over a wide area, which allow analysts to work out the direction from which a wave arrived and to pull faint signals out of background noise. An underground explosion and an earthquake both shake the ground, so the central scientific problem is #discrimination. Several classical criteria are used. An explosion is a sudden outward push from a point, so it generates strong compressional waves and relatively weak shear and surface waves. An earthquake, which is a slip on a fault, generates the opposite pattern. The ratio between body wave magnitude and surface wave magnitude therefore separates the two populations quite well for medium sized events. Depth is another clue: an event ten kilometres beneath the surface is not a bomb, because nobody can drill that deep for a test. Location is a third: an event inside a known test site is treated with more suspicion than one on a plate boundary. None of these criteria is perfect, and all of them weaken as events get smaller. This is where recent research has concentrated. Elkhouly and Ali (2024) compare classical spectral and magnitude ratio methods against linear and nonlinear machine learning classifiers, and find that learned models can add discriminating power. Steinberg et al. (2024) apply deep neural network denoising to regional waveforms recorded near the North Korean test site, and show that removing noise can reveal smaller events, while warning that the method can also create artefacts and must be used with care. Arora et al. (2024) tackle a different problem: large earthquakes produce long trains of scattered energy, and automatic systems sometimes mistake that energy for separate events, so their improved Bayesian model suppresses these false detections without losing genuine ones. Alamneh (2024) checks the quality of the Vienna centre's Reviewed Event Bulletin against the independent bulletin of the International Seismological Centre and finds a high rate of agreement with only small differences in location. Taken together, these papers describe a system that is accurate, and getting more accurate, at exactly the point where accuracy is hardest: the small event. 7.2 Hydroacoustic monitoring Sound travels remarkably well in the ocean. A layer of water at intermediate depth, known as the sound channel, traps acoustic energy and lets it propagate for thousands of kilometres with little loss. This is why only eleven #hydroacoustic stations are needed to cover the world's oceans. Six are hydrophone stations with underwater microphones, and five are seismic stations on islands or coasts that pick up the converted signal where ocean waves meet rock. The network has gaps, because continents block sound. Bittner et al. (2024) show that coastal seismic and infrasound stations can detect hydroacoustic signals and can therefore help fill those gaps, which is a cheap way of improving coverage without building new deep ocean infrastructure. 7.3 Infrasound monitoring Infrasound is sound below the range of human hearing. Large atmospheric explosions produce it, and so do volcanoes, meteors, rocket launches, and severe storms. Sixty #infrasound stations, each an array of sensitive barometers, listen for it. The scientific difficulty here is that the atmosphere is not a stable medium. Winds in the stratosphere change with season and can carry a signal towards one station and away from another. Listowski et al. (2024) study how internal gravity waves in the stratosphere disturb infrasound propagation, using high resolution simulations, and find measurable and hemisphere dependent effects on transmission loss. This kind of work is essential, because a station that cannot hear a signal is worse than useless if analysts do not know why. De Ceu-Jesus et al. (2024) take a different approach and show that a low cost portable array, deployed during a volcanic crisis in the Azores, can usefully supplement a permanent station on a neighbouring island. 7.4 Radionuclide monitoring The other three networks tell you that something exploded. The #radionuclide_network tells you whether it was nuclear. Eighty stations sample the air for radioactive particles, and about half of them also detect noble gases, particularly the isotopes of xenon. Sixteen laboratories provide confirmatory analysis. Xenon is the key. Fission produces xenon isotopes, xenon is chemically inert, and a gas that does not react with rock has a good chance of leaking to the surface even from a well contained underground shot. The ratio between different xenon isotopes can, in favourable conditions, distinguish a bomb from a reactor release. But the background is noisy. Medical isotope production facilities release xenon routinely, and so do power reactors. Kijima et al. (2024) develop a Monte Carlo method to estimate the statistical distribution of background xenon at each station, using known emissions and atmospheric transport modelling, which allows analysts to say how unusual a given detection really is. Kalinowski and colleagues have shown that spent fuel reprocessing is a small but non-negligible additional source. Goodwin et al. (2024) characterise the radionuclide fingerprint of an advanced gas-cooled reactor, again to help separate civil signals from weapons signals. Furuno et al. (2024) explain occasional caesium 137 detections in East Asia not as fresh contamination but as old fallout resuspended in Asian desert dust. Two lines of work push the technology forward. Adhikari et al. (2024) model whether argon 37, produced when neutrons interact with calcium in rock, could be detected at existing stations after an underground explosion, and treat it as a possible supplement to xenon monitoring. Pazdniakou et al. (2024) look at unconventional tracers such as water vapour and carbon dioxide emitted from a test site, and ask whether satellite imagery could see them. Azimi et al. (2024) and Heranudin (2024) work on the detectors themselves, improving energy resolution and absolute activity measurement. 7.5 The International Data Centre All of this data flows to Vienna. The #international_data_centre processes it automatically, produces lists of detected signals, associates signals from different stations into events, locates those events, and screens them against criteria that separate probable natural or industrial phenomena from events of possible interest. Human analysts then review the automatic output and produce the Reviewed Event Bulletin. Products are distributed to states signatories, who make the final judgement themselves. The centre does not declare that a treaty violation has occurred. That is a political decision reserved to states. Recent work has begun to test the whole pipeline rather than its parts. Berg et al. (2025) describe the Dynamic Network Experiments, in which four United States national laboratories pushed real multi-sensor data and synthetic radionuclide signatures through an end to end analysis chain to quantify how much advanced methods and better computing actually improve monitoring performance. This kind of systems level evaluation is new and valuable, because gains in one sensor are worth little if the pipeline cannot use them. 7.6 On-site inspection The final element is the most intrusive and the only one that is currently unavailable. If the monitoring data suggest that a nuclear explosion has occurred on the territory of a state party, another state party may request an #on_site_inspection. An inspection team may search an area of up to a thousand square kilometres, using visual observation, overflight, radiation measurement, passive and active seismic survey, magnetic and gravitational field mapping, ground penetrating radar, and drilling to collect samples. Because the treaty has not entered into force, no inspection can legally be requested. The Preparatory Commission has therefore spent two decades preparing for a capability it cannot yet use. Surrogate inspectors are trained, equipment is tested, and large integrated field exercises simulate the whole process. Research continues in parallel. Li et al. (2024) present a transportable active seismic survey system built around optical fibre sensors, designed to be used by an inspection team. Hafner et al. (2025) test a resonance seismometry technique, which uses the way seismic waves are disturbed by an underground cavity, on data from the Nevada site, and identify an anomaly near the location of a 1965 explosion. Wickramasinghe and De Silva (2024) point out that most inspection methods were designed with open, arid landscapes in mind, and describe the practical problems of working in tropical mountains with dense vegetation, heavy rainfall, poor satellite reception, and disease risk. Thu Zar et al. (2024) argue that inspectors could learn from the radiological protection experience gained after Chernobyl and Fukushima, where a process of shared expertise between specialists and affected populations helped communities make informed decisions. The gap between preparation and permission is the sharpest illustration of the treaty's condition. A complete inspection regime exists on paper and in training, and it cannot be used. 8. Civil and Scientific Applications A system built to detect bombs turns out to be extremely useful for other things, and this dual value is an underrated part of the treaty's political survival. Seismic data from the network feed earthquake research and, through arrangements with regional warning centres, contribute to #tsunami_warning. Hydroacoustic stations record whale calls, ice breaking off Antarctic shelves, and undersea landslides. Infrasound stations record meteor entries and volcanic eruptions. Radionuclide stations tracked the plume from the Fukushima accident in 2011 and have been used to study atmospheric transport. The eruption of Hunga Tonga-Hunga Ha'apai in January 2022 was recorded across all the waveform networks and produced an atmospheric pressure wave that travelled around the globe several times. Le Bras et al. (2025) compare the timing of these passages with historical records from the 1883 Krakatoa eruption and find differences in wave speed that may reflect changes in the atmosphere over the intervening period. Work of this kind is pure earth science, made possible by a #dual_use_infrastructure that exists for reasons of security. Oancea et al. (2024) describe a related contribution to the historical record: a modernised repository of waveforms and source parameters covering all two thousand one hundred and fifty seven known nuclear explosions between 1945 and 2017. That archive is a research resource, a verification benchmark, and, quietly, a memorial. The practical lesson for policy is straightforward. States that gain scientific and civil benefit from the network have a reason to keep funding it even when they are not willing to ratify the treaty. This is one of the mechanisms that has allowed the system to survive its own legal incompleteness. 9. The Compliance Record 9.1 North Korea The only state to have tested nuclear weapons in the twenty first century is the Democratic People's Republic of Korea, which conducted six announced explosions between 2006 and 2017 at the Punggye-ri site under Mount Mantap. Every one of them was detected by the International Monitoring System, located, and reported to states signatories within hours. The 2006 test, whose yield was under one kiloton, was detected even though it was small and well below the level that sceptics in the 1990s had said would be undetectable. These events have become the field's most studied natural experiment. Ziolkowski (2024) applies Bayesian inversion to seismograms from a single station to estimate source parameters and yields for tests two through six, and concludes that the site was already badly damaged by the explosions before its publicised demolition in 2018. Guilhem Trilla and Cano (2024) test a rapid moment tensor inversion method on the same events and on nearby natural earthquakes, and show that it can characterise both quickly and reliably. Berezina et al. (2024) analyse waveforms from the Lop Nur area recorded at Central Asian stations to compare explosion and earthquake signatures. The #dprk_tests are the clearest available evidence that the verification regime works. They are also evidence that verification alone does not prevent testing. Detection is not deterrence unless a political consequence follows detection. 9.2 The zero yield question The most difficult compliance issue is not a detected explosion but an alleged one. The treaty bans nuclear explosions without defining the term. The United States, the United Kingdom, and France interpret the ban as a #zero_yield standard: no self-sustaining supercritical chain reaction of any kind, however small. Under that reading, subcritical experiments, in which chemical explosives compress plutonium without producing a chain reaction, are permitted, and hydronuclear experiments, which produce a very small chain reaction, are not. American government reports since 2019 have expressed concern that Russia has conducted experiments producing more than zero yield, and have raised questions about transparency at Chinese facilities. Russia and China deny this. The problem is that the monitoring network, sensitive as it is, was designed for events measured in fractions of a kiloton, not for events measured in grams of explosive equivalent. A very small supercritical experiment conducted deep underground inside an existing tunnel complex may simply not produce a signal that any station can hear. This produces an uncomfortable but honest conclusion. For the great powers, at the very lowest yields, verification cannot fully settle the question, and the argument therefore becomes political rather than technical. Some analysts propose that the five recognised nuclear weapon states negotiate #transparency_measures at their former test sites, such as agreed instrumentation or reciprocal visits, precisely because the monitoring system alone cannot close this gap. 9.3 What the record shows Since 1996, the pattern is striking. Every state that has signed or ratified has refrained from testing. India and Pakistan, which have never signed, tested in 1998 and have observed unilateral moratoriums since. Israel has never openly tested. Only North Korea has broken the pattern, and it stopped in 2017. Whatever the treaty's legal status, the practice of not testing is close to universal. 10. Why Nine States Have Not Ratified 10.1 The United States President Clinton signed the treaty in 1996 and sent it to the Senate, which voted in October 1999. The treaty failed, with forty eight votes in favour and fifty one against, well short of the two thirds required. Three arguments carried the day: that the arsenal could not be certified as reliable indefinitely without testing, that the treaty could not be verified, and that a ban would constrain the United States more than its adversaries. All three arguments have weakened with time. The weapons laboratories have certified the stockpile annually for more than three decades without explosive testing, and successive laboratory directors and administrators of the National Nuclear Security Administration have stated that testing is not needed for reliability. The monitoring system has proved far more capable than expected. And the asymmetry argument runs the other way: the United States conducted more than a thousand tests, more than any other state, and holds the deepest archive of test data and the most advanced simulation capability. A world without testing preserves that advantage. A world with testing lets others catch up. Despite this, the treaty has never been returned to the Senate for a vote, and in 2025 the administration stated that it would not seek Senate advice and consent. In late October 2025, the President announced on social media that he had instructed the Department of War to begin testing nuclear weapons on an equal basis with other countries, immediately. As of mid 2026 no full yield explosive test had been conducted, and officials suggested that any activity would focus on subcritical or hydrodynamic experiments. The ambiguity of the announcement, and the gap between statement and action, is itself analytically important: it shows that the #testing_moratorium now depends on interpretation of a political statement rather than on a legal barrier. 10.2 China China signed in 1996 and has consistently said it supports the treaty and intends to ratify. It has not done so. The common explanation is that Beijing is waiting for Washington, and does not wish to accept a legal obligation that its principal rival has declined. China conducted its last test in 1996 and maintains a moratorium. Recent American reports have questioned the transparency of activity at Lop Nur, an accusation Beijing rejects. 10.3 India and Pakistan India refused the treaty in 1996 and has never signed. Its objections were both procedural and substantive: the entry into force clause was seen as coercive, and the treaty was seen as freezing a discriminatory order in which the existing nuclear powers keep their weapons and improve them by simulation while others are denied the option of testing. India tested in 1998 and has since declared a #voluntary_moratorium. Pakistan tested in response in 1998 and has said it will not test first if India does not, effectively tying its position to India's. 10.4 The Middle East cluster Egypt, Iran, and Israel form a linked set. Egypt has said it will not ratify while Israel remains outside the Nuclear Non-Proliferation Treaty and while progress toward a Middle East zone free of weapons of mass destruction is blocked. Israel has signed, participates in the monitoring system, and has said its ratification depends on the regional security environment. Iran has signed but not ratified, and its position is entangled with the wider dispute over its nuclear programme. This cluster illustrates a general feature of arms control: treaties fail not only for global reasons but because they are captured by #regional_security_dilemmas. 10.5 North Korea North Korea has never signed and is the only state to have tested since 1998. Its programme is now embedded in national law and doctrine, and there is no realistic prospect of ratification without a fundamental change in the security relationship on the Korean peninsula. 10.6 Russia Russia signed in 1996 and ratified in 2000. In October 2023, following remarks by President Putin, the Duma voted unanimously to revoke the ratification, and the law was signed in November 2023. Moscow's stated justification was that it was mirroring the position of the United States, which had signed but never ratified, and it said it would remain a signatory and continue to cooperate with the monitoring system. The legal analysis of this move is instructive. The withdrawal clause in Article IX cannot apply, because that clause, like the rest of the operative treaty, has no legal force until entry into force. Russia therefore acted under the general law of treaties, withdrawing an instrument of ratification for a treaty not yet in force. Under Article 18 of the Vienna Convention on the Law of Treaties, a state that has signed a treaty is obliged to refrain from acts that would defeat its object and purpose. Russia remains a signatory, so that obligation continues to bind it. In 2016 the five recognised nuclear weapon states jointly affirmed, and the United Nations Security Council endorsed, that a nuclear test would defeat the CTBT's object and purpose. On this reading, the #russian_deratification changed Russia's formal status without changing its substantive legal obligation not to test. The practical effect was political rather than legal. It removed a ratification from the Annex 2 count, raising the number of outstanding states from eight to nine, and it signalled that the treaty is now a bargaining chip in a wider confrontation. 11. The Treaty as a Norm If the CTBT has never entered into force, why has almost nobody tested for thirty years? This is the central puzzle, and the answer is usually given in the language of norms. A norm is a shared expectation about appropriate behaviour. Norms can exist without law, and they can outlive law. The claim in the literature is that the CTBT, together with the moratoria that preceded and accompanied it, created a #nuclear_testing_taboo: a widely held view that testing is not simply prohibited but disreputable, an act that marks a state as an outlier and invites diplomatic cost. Several mechanisms support this norm. The first is the monitoring system, which makes secrecy impossible for any test of meaningful size and therefore guarantees that the reputational cost will be incurred. The second is the near universality of signature, which means that a testing state faces almost the whole international community rather than a bloc. The third is the practical calculation described above: the states with the largest test archives have the most to lose from a return to testing, because testing helps rivals more than it helps them. The fourth is domestic, since testing carries public health, environmental, and political costs in the states that would host it. But norms are not self-enforcing, and the analytical literature on the wider non-proliferation regime is cautious. Gibbons and Herzog (2022) show that regimes which look durable can be strained badly by shifts in the distribution of power, and Bollfrass and Herzog (2022) trace how the war in Ukraine damaged the assumptions on which the nuclear order rested. Applied to the test ban, the argument is uncomfortable. A norm sustained by moratoria is a norm sustained by unilateral policy decisions, and unilateral policy decisions can be reversed by a single leader, in a single statement, without any legal process at all. The events of October 2025 demonstrate exactly that vulnerability. There is a further legal question that students should be aware of, even though it is unsettled. If almost every state has refrained from testing for three decades, and if states justify that restraint by reference to a shared obligation rather than mere convenience, has a rule of #customary_international_law formed? Some scholars argue yes, at least in respect of atmospheric testing. Others argue that the persistent objection of a few states, and the continued insistence by nuclear powers that they retain the option, prevents crystallisation. The debate matters practically, because a customary rule would bind even non-signatories. 12. The CTBT in the Wider Nuclear Order The test ban does not stand alone. It sits inside an architecture of overlapping agreements, and it is worth setting out those relationships clearly. The Treaty on the Non-Proliferation of Nuclear Weapons of 1968 is the foundation. Non-nuclear weapon states accept a ban on acquiring weapons; nuclear weapon states accept an obligation to pursue disarmament in good faith. The CTBT has always been treated as a #disarmament_deliverable within that bargain. It was promised at the 1995 conference that indefinitely extended the NPT, and it is invoked at every review conference as a test of whether the nuclear weapon states keep their side of the deal. When the eleventh review conference in 2026 again failed to reach a consensus final document, with testing among the contested issues, the damage to that bargain was real. The Treaty on the Prohibition of Nuclear Weapons, which entered into force in 2021, prohibits nuclear weapons outright for its parties. It has been ratified largely by non-nuclear weapon states. Its relationship to the older architecture is debated: some see it as a reinforcement of the disarmament norm, others as a source of division between states that rely on nuclear deterrence and states that reject it (Borrie 2021; Maitre 2025). For the test ban, the practical consequence is that the CTBT remains the one instrument on which both camps can agree, and this makes it valuable diplomatic ground. A proposed Fissile Material Cut-off Treaty, which would stop the production of fissile material for weapons, has been stuck in the Conference on Disarmament for decades. The pairing is often described as: the CTBT would stop qualitative improvement, and a cut-off treaty would stop quantitative growth. Neither is complete. Finally, regional #nuclear_weapon_free_zones in Latin America, the South Pacific, Southeast Asia, Africa, and Central Asia all prohibit testing within their territories, and their members are among the strongest supporters of the CTBT. The Central Asian zone is centred on Kazakhstan, the state that lived with Semipalatinsk, and that fact is not a coincidence. 13. Humanitarian and Environmental Dimensions Any account of the test ban that stays at the level of treaty law and sensor networks misses why it exists. Testing was not a victimless technical activity. At Semipalatinsk in Kazakhstan, hundreds of explosions were conducted near inhabited areas, and the health consequences for local populations are still being documented. In the Marshall Islands, whole communities were displaced, some permanently, and the 1954 Castle Bravo test deposited fallout on inhabited atolls. In French Polynesia, Algeria, Australia, Nevada, and Xinjiang, similar stories exist with local variations. The people affected were often colonised, poor, indigenous, or simply distant from the capitals that made the decisions. These histories produced political movements, and those movements produced the treaty. The Nevada Semipalatinsk movement was central to the Soviet decision to stop testing in 1990. Pacific campaigns shaped Australian and New Zealand policy. The #humanitarian_argument is not a decorative addition to the arms control case. It is the reason the arms control case gained political traction. It also has a forward looking dimension. Underground testing is safer than atmospheric testing, but it is not safe. Containment can fail: the Baneberry test in Nevada in 1970 vented a radioactive plume that exposed workers. Any resumption of testing in Nevada would take place near a metropolitan area far larger than the one that existed in 1992. And the states that would test are, without exception, states whose own citizens would bear some of the risk. 14. Current Challenges Five challenges define the treaty's present situation. The first is the #entry_into_force_deadlock. Article XIV gives forty four states a veto, and the incentives of the holdouts are locked together: China waits for the United States, Pakistan waits for India, Egypt waits for Israel, and each can point to another's inaction as justification. Nothing in the treaty's design breaks this cycle. The second is the erosion of the moratorium. The statements of October and November 2025, and the Russian response that it would answer in kind if others tested, moved the question from theoretical to live. The moratorium is now the only barrier, and it is a barrier made of political will. The third is the #definitional_gap at very low yields. Without an agreed definition of a nuclear explosion, and without the ability to verify below a certain threshold, accusations of cheating cannot be settled by data. This creates a self-reinforcing spiral in which each side's suspicion justifies its own preparations. The fourth is the collapse of the surrounding architecture. New START, the last bilateral treaty limiting American and Russian strategic warheads, has reached the end of its life, and the review process of the NPT has failed repeatedly to reach consensus. A test ban that is not in force is more exposed when everything around it is weakening. The fifth is technological. Advances in simulation, in inertial confinement fusion facilities, and in artificial intelligence raise questions about how much weapons development can be done without any explosive testing at all, and therefore about how much a test ban actually constrains a sophisticated state. This is a genuine limit on what the treaty can achieve, and honest analysis should say so. A test ban prevents the confident development of radically new designs by states without extensive test archives. It does not freeze weapons science in states that already have those archives. 15. Pathways Forward The literature and the diplomatic record suggest several possible responses, and students should be able to weigh them critically rather than simply list them. The first is renewed pressure for ratification. This is the official strategy of the Article XIV conferences. Its record is poor, but its logic is sound, and it costs little. The second is #confidence_building at test sites. If the core dispute concerns very low yield activity that the monitoring system cannot resolve, then the answer must come from transparency rather than from sensors. Proposals include reciprocal visits, agreed instrumentation at former test sites, and joint technical talks among the five recognised nuclear weapon states. The value of this approach is that it addresses the actual point of failure. Its weakness is that it requires exactly the trust that is currently absent. The third is a formal political commitment to the moratorium. A joint declaration by the nuclear weapon states, or a Security Council resolution reaffirming that a test would defeat the object and purpose of the treaty, would not have the force of the treaty, but it would raise the political cost of a first test. The fourth is the strengthening of the #verification_regime itself. Completing the remaining stations, pushing detection thresholds lower, and improving on-site inspection readiness all reduce the space in which a state could test undetected, and all can be done without entry into force. Sato's (2021) argument is that this path should be pursued on its own terms, and not held hostage to the ratification question. The fifth, more radical, is to revisit Article XIV. Some have proposed an alternative entry into force mechanism, or a separate agreement among willing states. Both would require reopening a negotiation that took years to conclude, and both risk destroying the near universal signature that is the treaty's greatest asset. Most analysts regard this as a last resort. 16. A Research Agenda for Students The CTBT is unusually rich as a subject for student research, because it sits at the intersection of law, physics, politics, and history. Several questions are open and tractable. In the technical domain: how much do machine learning classifiers actually improve #seismic_discrimination when tested on genuinely independent data, rather than on the same events used for training? Where are the remaining coverage gaps in the hydroacoustic and infrasound networks, and what would close them at least cost? How well can radionuclide background be modelled at stations near medical isotope production facilities? In the legal domain: what exactly does the obligation not to defeat the object and purpose of a signed treaty require in practice, and would a subcritical experiment, a hydronuclear experiment, or a very low yield test each breach it differently? Has a customary prohibition on testing formed, and what evidence would settle the question? In the political domain: what would actually move one of the nine holdout states? Under what conditions do states ratify treaties they have signed? Public opinion research, of the kind conducted on the prohibition treaty in Japan (Kanno and Miwa 2024), could be extended to testing, and would tell us whether domestic publics constrain leaders on this issue at all. In the historical and humanitarian domain: the archives of the test era continue to open, and the health and environmental record of testing in Kazakhstan, the Pacific, Algeria, and Xinjiang remains incompletely written. The waveform repository described by Oancea et al. (2024) is a data resource that most students have never heard of, and it is available for analysis. The most important lesson for a student researcher is one about disciplinary humility. The technical papers cannot tell you whether the treaty will enter into force. The political papers cannot tell you whether a one tonne yield explosion in a deep tunnel would be detected. Good work on the CTBT requires reading across the boundary. 17. Conclusion The Comprehensive Nuclear-Test-Ban Treaty is a paradox that has lasted thirty years. It is not in force, and yet it has stopped nuclear testing almost everywhere. It has no legal power to inspect, and yet it operates a global sensor network of extraordinary sensitivity. It has been signed by one hundred and eighty seven states, and blocked by nine. Three conclusions follow from the review presented here. First, the technical case against the treaty has collapsed. The argument of 1999, that a comprehensive ban could not be verified, is not defensible today. The system detected every North Korean test, including the smallest, and the peer reviewed literature from 2021 to 2025 shows steady improvement in detection, location, discrimination, and background estimation across all four technologies. What remains unverifiable is a narrow class of extremely low yield experiments, and that residue is a problem for #transparency, not a problem for seismology. Second, the political case is now weaker than it has been at any point since 1996. The Russian withdrawal of ratification in 2023, the failure of the review process of the wider non-proliferation regime, the expiry of the last bilateral limits on strategic arsenals, and the American statements of late 2025 have together turned the moratorium from a stable feature of the landscape into an open question. The treaty's protective effect was always borrowed from the norm. The norm is now under strain. Third, the gap between these two facts is the heart of the matter. The CTBT has built a machine that can enforce a rule that does not legally exist. Whether that machine is ever switched on fully depends on decisions that have nothing to do with science. For thirty years, the world has enjoyed the benefits of a test ban without paying the political price of ratifying one. That arrangement has been quietly successful and is now visibly fragile. The ban on testing was not achieved by governments alone. It was demanded by people who lived downwind. If the norm is to survive the current period, the argument will have to be made again, in the same terms: that a nuclear explosion, wherever it happens and whatever it is called, is a harm to everyone, and that the world does not need to hear one again to know what it does. Hashtags #CTBT_1996 #nuclear_test_ban #arms_control #nonproliferation #disarmament #CTBTO #verification #international_monitoring_system #on_site_inspection #nuclear_policy #treaty_law #global_security #seismology #peace_studies #international_relations References Adhikari, P., Gordon, E. M., Shah, K. A., et al. (2024). Quantifying the potential of argon detection capabilities for nuclear explosion monitoring. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03489-x Alamneh, F. K. (2024). Quality assessment of the reviewed event bulletin through comparison with the ISC bulletin for the month of October 2020 and summary results of earlier assessments. Pure and Applied Geophysics, 181, 815-827. https://doi.org/10.1007/s00024-024-03430-2 Arora, N. S., Ali, S. M., Shashkin, A., et al. (2024). Suppressing coda events with a Bayesian model of global scale seismology. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03574-1 Azimi, S. A., Afarideh, H., and Chai, J. S. (2024). Compact beta-gamma radioxenon detection system with improved energy resolution using PIPS detector. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03487-z Baron, J., Gibbons, R. D., and Herzog, S. (2022). Antinormative messaging, group cues, and the nuclear ban treaty. Journal of Politics, 84(1), 591-596. Berezina, A., Sokolova, I., Kopnichev, Y., et al. (2024). Features of the waveforms of nuclear explosions and earthquakes from the Lop Nor test site area according to the data of seismic stations in Central Asia. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03482-4 Berg, E. M., Carmichael, J. D., Young, C. J., et al. (2025). The dynamic networks experiments: virtual experiments to quantify gains in nuclear explosion monitoring. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-025-03808-w Bittner, P., Ali, S. M., Kasmi, A., et al. (2024). Hydroacoustic signals detected at coastal international monitoring system stations. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03585-y Bollfrass, A. K., and Herzog, S. (2022). The war in Ukraine and global nuclear order. Survival, 64(4), 7-32. Borrie, J. (2021). An introduction to implementing the Treaty on the Prohibition of Nuclear Weapons. Journal for Peace and Nuclear Disarmament, 4(1), 1-12. https://doi.org/10.1080/25751654.2021.1947460 de Ceu-Jesus, M., Belli, G., Gheri, D., et al. (2024). The use of a low-cost, small-aperture array as an auxiliary tool to improve infrasound monitoring in the Azores region. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03469-1 Elkhouly, S. H., and Ali, G. (2024). Seismic discrimination between nuclear explosions and natural earthquakes using multi-machine learning techniques. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03463-7 Fetter, S., and Sankaran, J. (2025). Emerging technologies and challenges to nuclear stability. Journal of Strategic Studies, 48(2), 252-296. Findlay, T. (2022). Transforming nuclear safeguards culture: the IAEA, Iraq, and the future of nonproliferation. MIT Press. Fuhrmann, M. (2024). Influence without arms: the new logic of nuclear deterrence. Cambridge University Press. Furuno, A., Ohmori, R., Tateoka, H., et al. (2024). Assessment of caesium-137 detections at CTBTO radionuclide monitoring stations in East Asia and their relationship to Asian dust dispersion. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03620-y Geist, E. (2023). Deterrence under uncertainty: artificial intelligence and nuclear warfare. Oxford University Press. Gibbons, R. D., and Herzog, S. (2022). Durable institution under fire? The NPT confronts emerging multipolarity. Contemporary Security Policy, 43(1), 50-79. Goodwin, M. A., Petts, A., Milbrath, B. D., et al. (2024). Characterising the radionuclide fingerprint of an advanced gas-cooled nuclear power reactor. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03488-y Guilhem Trilla, A., and Cano, Y. (2024). Using moment tensor inversions for rapid seismic source detection and characterization: application to the North Korean nuclear tests. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03455-7 Hafner, R., Walter, M., and Joswig, M. (2025). Testing the onset delay method for resonance seismometry: the underground nuclear explosion Tiny Tot case study. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-025-03746-7 Heranudin. (2024). Development of a new 4 pi beta - 4 pi gamma detection system for absolute measurement of radionuclide activity. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03543-8 Hunt, J. R. (2022). The nuclear club: how America and the world policed the atom from Hiroshima to Vietnam. Stanford University Press. Kalinowski, M. B., and Mialle, P. (2021). Introduction to the topical issue on nuclear explosion monitoring and verification: scientific and technological advances. Pure and Applied Geophysics, 178(7), 2397-2401. https://doi.org/10.1007/s00024-021-02783-2 Kalinowski, M. B., Sarid, E., Mialle, P., et al. (2023). Innovation in technology and scientific methods for nuclear explosion monitoring and verification: introduction. Pure and Applied Geophysics, 180, 1227-1234. https://doi.org/10.1007/s00024-023-03292-0 Kanno, T., and Miwa, H. (2024). Military threats, international pressure, and public opinion on the nuclear ban treaty in Japan. Journal for Peace and Nuclear Disarmament, 7(2), 446-467. https://doi.org/10.1080/25751654.2024.2387896 Kijima, Y., Schoemaker, R., Tipka, A., et al. (2024). Investigation of radioxenon probability density functions at IMS radionuclide stations using a Monte Carlo method for background estimation. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03541-w Klaus, L., Schwardt, M., Pilger, C., et al. (2024). Linking seismic measurements to the international system of units. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03493-1 Le Bras, R., Bittner, P., Kusmierczyk-Michulec, J., et al. (2025). Timing analysis of the multiple passages of the pressure wave generated by the 2022 Hunga Tonga-Hunga Ha'apai eruption and comparison with the 1883 Krakatoa pressure wave. Pure and Applied Geophysics, 182, 1-28. https://doi.org/10.1007/s00024-024-03507-y Li, Q., Li, Z., Wang, Z., et al. (2024). Optical fiber sensor based transportable active seismic survey system solution to OSI. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03539-4 Listowski, C., Stephan, C. C., Le Pichon, A., et al. (2024). Stratospheric gravity waves impact on infrasound transmission losses across the international monitoring system. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03467-3 Maitre, E. (2025). Abolishing nuclear weapons at a time of nuclear revival: prospects for the TPNW. Journal for Peace and Nuclear Disarmament, 8(1), 169-191. https://doi.org/10.1080/25751654.2025.2516369 Mialle, P., and Kalinowski, M. B. (2025). Nuclear explosion monitoring and verification: science and technology to tackle global challenges: an introduction. Pure and Applied Geophysics, 182(12), 4845-4852. https://doi.org/10.1007/s00024-025-03890-0 Mulwa, J. K. (2024). Systematic seismic events discrimination methods at the Kenya National Data Centre (N090). Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03458-4 Oancea, V., Kung, Y. L., Murphy, J. R., et al. (2024). Waveforms from nuclear explosions (WFNE) repository. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03536-7 Pazdniakou, A., Mourzenko, V., Thovert, J. F., et al. (2024). Enhancing detection of underground nuclear tests with unconventional tracers. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03595-w Sato, M. (2021). Advancing nuclear test verification without entry into force of the CTBT. Journal for Peace and Nuclear Disarmament, 4(2), 251-267. https://doi.org/10.1080/25751654.2021.1993643 Simon, R., King, J. G., Moffat, L. C., et al. (2024). Relocation of the 3 April 2017 Moiyabana earthquake in Botswana using seismological data from a local network of autonomously recording seismographs and international monitoring system stations. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03435-x Steinberg, A., Gaebler, P., Hartmann, G., et al. (2024). Deep neural networks based denoising of regional seismic waveforms and impact on analysis of North Korean nuclear tests. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03491-3 Thu Zar, W., Lochard, J., Kalinowski, M. B., et al. (2024). What on-site inspectors under the Comprehensive Nuclear-Test-Ban Treaty can learn from the co-expertise process experiences implemented after the Chernobyl and Fukushima nuclear power plant accidents. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03501-4 Vilmer, J. B. J. (2022). The forever-emerging norm of banning nuclear weapons. Journal of Strategic Studies, 45(3), 487-513. https://doi.org/10.1080/01402390.2020.1770732 Wickramasinghe, W. A. G. K., and De Silva, S. N. (2024). Additional challenges for on-site inspection (OSI) in tropics. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03610-0 Ziolkowski, A. (2024). Bayesian inversion of MDJ seismograms for source parameters and yields of DPRK underground nuclear explosions 2 to 6. Pure and Applied Geophysics. https://doi.org/10.1007/s00024-024-03454-8