When Nuclear Threats Stop Working: Why That Could Make the World More Dangerous

[DEFCON Warning System]

Russia’s latest warnings over Kaliningrad present a familiar problem in an increasingly pointed form. In late September, Russian diplomatic statements warned that attempts to isolate the exclave could bring a nuclear response. NATO rejected the suggestion that it was threatening Russian territory. On 1 October, Vladimir Putin said a direct attack would put the use of Russia’s entire arsenal on the agenda.[1][2]

The response was almost as significant as the warning. Asked about the Russian message on 30 September, NATO Secretary General Mark Rutte said, “I don’t take this that seriously”, while also explaining that the alliance had replied.[3]

That exchange does not prove that NATO has become indifferent to nuclear danger. Public confidence can itself be strategic messaging. But it raises a consequential question: what happens when warnings intended to inspire caution increasingly invite dismissal?

The evidence supports a qualified answer. Nuclear threats can fail to prevent particular actions without losing all deterrent value. Repeated unsuccessful warnings may encourage further challenges, but there is no established rule that repetition produces indifference, or that indifference necessarily increases nuclear risk. Indeed, recent public-opinion research provides evidence that Russian threats still influence attitudes towards military assistance to Ukraine.[4]

The danger is more specific: governments may correctly conclude that one nuclear warning was coercive, then incorrectly assume that another warning—issued under materially different circumstances—must also be empty.

What, exactly, has stopped working?

Deterrence seeks to prevent an action: attack us, and you may face unacceptable retaliation. Compellence seeks to make an adversary act: withdraw, surrender territory, or change policy, or face punishment. Both belong to the broader practice of coercion, but they impose different demands on credibility.[5]

Threatening nuclear retaliation against an attack that could destroy a state is comparatively understandable. Threatening nuclear war to obtain a limited diplomatic concession raises an obvious question: would the threatening government really accept the consequences?

Todd Sechser and Matthew Fuhrmann’s historical and quantitative research challenges the assumption that possessing nuclear weapons provides a general advantage in compellent diplomacy. Their findings support distinguishing the ability to protect vital interests from the ability to force others to comply with political demands. The wider scholarly debate remains contested, including over how coercive success should be measured.[6]

This distinction complicates claims that Russia’s nuclear threats have “failed”. A warning intended to discourage a particular weapons delivery can fail even while fear of escalation continues to discourage direct NATO–Russia combat.

NATO’s refusal to establish a no-fly zone over Ukraine in March 2022 illustrates the difference. Jens Stoltenberg explained that enforcement would require fighting Russian aircraft and could produce a much wider European war. This demonstrates restraint over direct intervention, although it cannot establish how much restraint resulted from nuclear weapons rather than conventional risks and other considerations.[7]

Nor does continued Western assistance prove that threats achieved nothing. Assistance might have arrived earlier, in greater quantities, or with fewer restrictions without escalation concerns. That counterfactual cannot be observed directly.

Losing confidence in nuclear rhetoric is therefore not necessarily the same as losing respect for nuclear capability.

Habituation, or sensible learning?

“Nuclear-threat fatigue” can describe several processes that should not be conflated.

People may become emotionally accustomed to repeated warnings. Governments may revise their estimates after threatened consequences fail to materialise. Military planners may identify options that appear to remain below an adversary’s threshold. Journalists may give familiar statements less attention.

Only the second and third necessarily involve a judgement about strategic credibility. Reduced anxiety, reduced coverage, and reduced deterrence are different outcomes.

Learning from restraint can be sensible. If a government repeatedly associates limited actions with catastrophic consequences, but responds through conventional or diplomatic means, its opponents gain evidence that its public language overstates its immediate willingness to escalate.

The analytical error comes when that lesson becomes universal. Restraint after a weapons shipment does not establish restraint after attacks on national leadership, nuclear command systems, or forces essential to retaliation. Different stakes can produce different decisions.

Even an apparently ignored warning is ambiguous. Perhaps the threatening state was bluffing. Perhaps its threshold was never reached. Perhaps private warnings, military precautions, or diplomatic intervention prevented escalation. Perhaps its leaders considered nuclear use and rejected it.

William Burns captured this tension in September 2024. The then CIA director said there had been a genuine risk of potential Russian tactical nuclear use in autumn 2022, while arguing that the West should not be unnecessarily intimidated. He also described conveying consequences directly to his Russian counterpart.[8]

That is evidence against treating every warning as theatre. It is also evidence against treating concern about nuclear escalation as a reason for automatic compliance.

 

[DEFCON Warning System]

Russia’s interests extend beyond its president

Kaliningrad matters partly because geography makes isolation a plausible Russian security concern. However, a Russian claim that NATO intends a blockade is not evidence that such a plan exists.

Russia’s November 2024 nuclear policy explicitly identifies attempts to isolate part of its territory, including blocking vital transport connections, as a military danger to be addressed through nuclear deterrence. That category is distinct from the document’s conditions permitting nuclear employment. Treating every listed danger as an automatic launch trigger misreads the policy.[9]

The document also allows nuclear use against conventional aggression posing a critical threat to Russian or Belarusian sovereignty or territorial integrity. It treats aggression by a non-nuclear state with a nuclear state’s participation or support as a joint attack. These provisions widen the declared circumstances of concern; they do not mandate nuclear use after every qualifying incident.[9]

Russia’s reliance on nuclear weapons is not solely a product of Putin’s personality. CNA research tracing Russian military debates identifies roots in the conventional weakness of the 1990s and subsequent efforts to integrate nuclear, conventional, and non-military instruments into escalation management. Demonstration, intimidation, and selective force occupy places within that thinking.[10]

Another Russian president would inherit the arsenal, military institutions, geography, and concern about the survivability of national command systems. Such continuities make it reasonable to expect some nuclear incentives to persist. They do not establish that another leader would make identical threats or accept identical risks.

Historical narratives also require careful treatment. Memories of invasion, the Soviet collapse, and lost international status may shape how elites interpret vulnerability. Their political significance depends on how leaders and institutions select and use those memories. They are not fixed national characteristics.

Putin’s personal role nevertheless matters. In 2023, Burns publicly assessed that a narrowing advisory circle and growing confidence in a historical mission had increased Putin’s appetite for risk. That is an informed external assessment, not direct access to his private thinking.[11]

The most consequential possibility is that a leadership group could confuse the survival of its government, its military campaign, or its prestige with the survival of Russia itself. Public evidence cannot establish when, or whether, that conflation would determine a nuclear decision.

History offers warnings, not a formula

The Cuban Missile Crisis shows both the power of fear and the limits of control.

In October 1962, Soviet missiles in Cuba brought the United States and Soviet Union into a direct confrontation. The settlement combined Soviet withdrawal with an American non-invasion assurance and a privately communicated understanding about removing American Jupiter missiles from Turkey. It was not simply a demonstration that the more threatening side prevailed.[12]

At sea, participants operated with much less information. Records and recollections assembled by the National Security Archive describe the dangerous encounter involving Soviet submarine B-59, its nuclear torpedo, and American anti-submarine forces. Vasily Arkhipov helped avert a potentially catastrophic decision amid confusion about American actions.[13]

Cuba is not evidence of nuclear-threat fatigue. It demonstrates something relevant to the proposed mechanism: leaders can intend controlled pressure while people closer to the confrontation experience something that looks like war.

Richard Nixon’s October 1969 nuclear alert illustrates another problem. The administration secretly increased military readiness to influence Soviet calculations and negotiations over Vietnam. The operation included nuclear-armed bomber activity, making it more than rhetoric. Archival research finds that the effort failed to produce the intended diplomatic breakthrough.[14]

A stronger signal therefore need not communicate more successfully. The recipient may notice military activity without accepting—or even correctly understanding—the intended demand.

India and Pakistan offer a further warning about learning contradictory lessons. After their 2001–02 confrontation, Indian and Pakistani leaders publicly advanced different explanations for why war had been avoided. Each could interpret the outcome as evidence that its own pressure or deterrence had worked.[15]

The subsequent Pakistani argument for short-range nuclear weapons is especially relevant. In a 2015 public discussion, former Strategic Plans Division chief Khalid Kidwai explained the Nasr system as a response to perceived Indian efforts to find room for conventional operations below Pakistan’s nuclear threshold.[16]

This is evidence of a state strengthening its nuclear options because it believed an adversary saw a gap in deterrence. It is not proof that repeated verbal threats caused that development.

Together, these cases support a limited conclusion: perceived failures of deterrence can encourage stronger signalling or altered capabilities, while both sides may leave a crisis more confident in incompatible assumptions.

Different states, different signals

The underlying problem extends beyond Russia, but its expression varies.

China’s November 2025 arms-control white paper reaffirmed its no-first-use policy and described its nuclear forces as serving self-defence. That is an authoritative statement of declared policy. It does not independently establish how Chinese leaders would interpret every possible wartime contingency.[17] It should also be noted that policies are flexible in a wartime environment.

The United States’ 2022 Nuclear Posture Review, by contrast, reserved nuclear use for extreme circumstances involving vital American, allied, or partner interests. This formulation illustrates how a state can signal restraint while retaining discretion over exceptional cases.[18]

India’s published 2003 position combined no first use with a reservation concerning major chemical or biological attacks. Pakistan has emphasised the need to deter conventional operations as well as nuclear attack.[19][16] These differences reflect particular security relationships and institutional choices, not immutable national temperaments.

North Korea’s 2022 nuclear law presents another model. Its stated conditions encompass attacks or imminent attacks against leadership and nuclear command, and it describes an automatic response if nuclear command and control is endangered. The declaration is evidence of the message Pyongyang wants opponents to receive; it is not verification of an automatic technical mechanism.[20]

The practical implication is that familiarity with a government’s threatening language cannot substitute for examining its vulnerabilities. A leadership that expects an attempt to remove it may interpret a conventional operation differently from an adversary that regards the same operation as limited.

Reputation matters, but reputation is not a single score. A government can be considered unreliable about minor threats and credible about defending its continued existence.

 

[DEFCON Warning System]

People interpret the danger

Nuclear decisions pass through human beings, institutions, and imperfect information.

Robert Jervis’s work on perception explains why decisionmakers do not simply update their beliefs neutrally as evidence arrives. Existing expectations influence what they notice, how they interpret ambiguity, and which historical analogies they consider relevant.[21]

Repeated restraint can reinforce the belief that an opponent always backs down. Repeated confrontation can reinforce the opposite belief: that every apparently limited action is part of a larger hostile design.

Fear, pride, anger, humiliation, nationalism, and ideological conviction may influence these judgements. Prestige and domestic pressure can make retreat politically costly. Groupthink and bureaucratic incentives can discourage advisers from challenging a preferred interpretation. Fatigue and incomplete information can narrow the options that receive serious consideration.

These are mechanisms to investigate, not diagnoses of particular leaders. Their effects are not uniformly escalatory. Fear can encourage compromise; professional military advice can restrain a political leader; concern for reputation can favour reliability and restraint.

The danger is greatest when these pressures produce mismatched perceptions. One leadership believes it is carefully testing an established boundary. The other believes the boundary has already been crossed and that further restraint will invite a larger attack.

Has the public stopped listening?

The evidence does not justify claiming that populations have generally become numb to nuclear threats.

A study published in 2026 by Julian Wucherpfennig and colleagues, using a German survey experiment with 3,048 respondents, found that more than 40 per cent considered Russian nuclear threats credible. Perceived credibility reduced support for supplying modern weapons to Ukraine. This suggests a possible route through which nuclear pressure can influence democratic policy without directly persuading government decisionmakers.[4]

It does not show that threats determine actual government policy, nor does it establish how the same individuals respond after years of repeated exposure.

Other research also challenges simple assumptions about normalisation. An experiment involving Russian respondents found greater aversion to nuclear strikes than to comparable conventional strikes, including opposition to a demonstrative nuclear explosion in an unpopulated area.[22]

Meanwhile, a 2026 review of post-Cold War research identified uneven geographical coverage and substantial gaps in knowledge about public responses to nuclear weapons. Its findings counsel particular caution about sweeping claims across societies or generations.[23]

Media repetition could reduce the novelty of a warning. Fatalism could make people disengage despite continued fear. Distance from Hiroshima, Nagasaki, and the Cold War could weaken historical familiarity. These are plausible explanations requiring better longitudinal evidence, rather than established descriptions of the public as a whole.

Their strategic effects could also differ. Reduced fear might weaken pressure for concessions. Fatalism might reduce demands for risk reduction. Persistent anxiety might increase support for restraint.

Public attitudes can shape electoral incentives and acceptable policy, but they do not translate automatically into military decisions. A population that ignores a headline may still oppose nuclear use; an outwardly unconcerned government may still be reviewing classified warnings carefully.

Why declining coercive power could improve stability

The strongest counterargument deserves more than a passing acknowledgement.

If every nuclear warning produces concessions, governments acquire an incentive to issue more warnings. Nuclear capability can then become protection for conventional aggression or a means of demanding political privileges.

Resistance can reduce that reward. When opponents continue limited policies without attacking a nuclear state’s survival, they may establish a more credible distinction between ordinary competition and genuinely vital interests.

Successful learning could therefore strengthen stability. Governments might make narrower threats, explain their interests more clearly, and place greater weight on observable behaviour. Their opponents might become less vulnerable to inflammatory language while remaining sensitive to actual changes in danger.

The coercive-diplomacy scholarship supports scepticism towards claims that nuclear possession can reliably compel concessions.[6] NATO’s early distinction between supporting Ukraine and directly enforcing a no-fly zone shows that assistance and escalation restraint can coexist.[7]

However, failed threats do not necessarily teach the threatening government to stop. Leaders may believe their message was insufficiently explicit, their preparations insufficiently visible, or their opponents insufficiently frightened.

The same unsuccessful threat can produce opposite lessons: one side learns that resistance is safe; the other learns that greater pressure is necessary.

 

[DEFCON Warning System]

The credibility trap

A credibility trap becomes possible when a government publicly commits itself to consequences it would prefer not to impose, then comes to believe that backing down threatens something more important than the original dispute.

It may respond with sharper language, exercises, visible deployments, or changes in readiness. Russian escalation-management literature and Pakistan’s explanation for developing Nasr make this general logic more than an abstract possibility.[10][16]

But there is no evidentiary basis for a mechanical progression from ignored warning to nuclear demonstration to nuclear attack. Leaders can redefine their objectives, retaliate conventionally, accept an unfavourable outcome, or seek a negotiated exit.

A nuclear demonstration would also create fresh uncertainty. It might frighten an adversary into restraint, strengthen resistance, or be interpreted as preparation for further use. No government can reliably dictate the political meaning of a detonation.

Another route to escalation does not require anyone to decide that credibility must be restored. James Acton’s research on nuclear–conventional “entanglement” examines capabilities serving both missions. A conventional attack on shared communications, sensors, or command infrastructure could be interpreted as an attempt to disable nuclear retaliation.[24]

Here, excessive confidence in earlier restraint becomes particularly dangerous. An attacker may think it is extending an established conventional campaign. The defender may think it is losing the ability to survive the next stage.

The central risk is thus not simply that an embarrassed leader might act rashly. It is that accumulated confidence, changing military circumstances, and incompatible interpretations can interact faster than governments revise their assumptions.

When words begin to mean something different

An explicit nuclear threat, an ambiguous reference to “all means”, and a reminder of nuclear capability communicate different degrees of specificity. None, alone, proves that employment preparations are under way.

A doctrinal revision is more consequential than a commentator’s statement because it formally changes declared policy. An exercise or deployment adds observable activity. Yet neither automatically demonstrates an intention to use nuclear weapons.

Assessment should ask whether behaviour is unusual for that force, whether several independent developments coincide, and whether the political context explains them.

Changes in readiness, dispersal, or movement deserve attention when they depart from established patterns. Movement of nuclear-capable systems is less informative if those systems also perform conventional missions. Activity involving nuclear units becomes more concerning when it resembles employment preparation rather than routine training.

Leadership relocation, changes to command arrangements, unusual military communications, and measures to preserve continuity of government may also matter. Their meaning depends on corroboration: protective actions can indicate fear of attack rather than plans to initiate one.

Historical alerts demonstrate that military preparation can be used for signalling, while research on entanglement shows why preparations or attacks can be misinterpreted.[14][24] The most significant development would be a convergence of political warnings, altered operational behaviour, and circumstances that leaders plausibly regard as threatening vital interests.

These are considerations for interpretation, not a countdown checklist. Public observers may lack decisive information. Absence of visible preparation is not proof of safety, just as an isolated movement is not proof of impending use.

The danger is learning the wrong lesson

The evidence supports the article’s premise conditionally. Repeated nuclear warnings can lose influence over particular decisions, and perceived gaps in deterrence can encourage stronger measures. But research does not establish a general causal chain from repetition, through habituation, to nuclear war.

Nor does it show that nuclear fear has simply disappeared. Threats can still affect public preferences, and governments can reject coercion while continuing to exercise significant restraint.

The decisive distinction is between learning that a particular demand lacks credibility and learning that nuclear escalation is never credible. The first can reduce the rewards of intimidation. The second can encourage actions whose significance the attacker does not understand.

A nuclear threat automatically believed can enable coercion. A nuclear threat automatically dismissed can enable miscalculation. Stability therefore depends on governments resisting demands without becoming prisoners of their own confidence: previous restraint must inform their judgement, but never prevent them from recognising when the stakes, preparations, and danger have genuinely changed.

 

[DEFCON Warning System]

Numbered endnotes

1. Associated Press, 30 September 2026, “Russia warns NATO of a possible nuclear response if Kaliningrad is cut off”. Reports the embassy statements, NATO’s confirmation of a written message, and its response. Russia’s allegation of blockade preparations is attributed, not independently established.
2. Reuters, 1 October 2026, “Putin Says World Is Living Through a Dangerous Moment, Warns West Not to Escalate”, republished by The Moscow Times. Used for the reported Valdai remarks; an accessible original Kremlin transcript was not located.
3. NATO, 30 September 2026, remarks by Secretary General Mark Rutte at the Euronews Defence & Space Summit. Official transcript, including the question about Russia’s message and Rutte’s reply.
4. Julian Wucherpfennig, Christian Gläßel, Felix Lemmer, and Dino Kolonić, “Weapons of Mass Deterrence: How Nuclear Threats Succeed”, Journal of Global Security Studies, 11(1), published online 17 April 2026. See also the researchers’ institutional account, Hertie School, 23 April 2026, “How nuclear threats succeed: Public opinion as a channel of coercion”. The accessible abstract and institutional summary support the findings reported here.
5. Thomas C. Schelling, Arms and Influence, Yale University Press, 1966. Foundational treatment of deterrence, compellence, bargaining, and the manipulation of risk.
6. Todd S. Sechser and Matthew Fuhrmann, Nuclear Weapons and Coercive Diplomacy, Cambridge University Press, 2017. For competing interpretations and methodological qualifications, see H-Diplo/ISSF Roundtable XI-15, 2019.
7. NATO, 4 March 2022, press conference by Secretary General Jens Stoltenberg following the extraordinary meeting of NATO Ministers of Foreign Affairs. Primary source for the explanation of why NATO rejected a no-fly zone.
8. CIA, 7 September 2024, transcript of William Burns and Richard Moore’s discussion with Financial Times editor Roula Khalaf, especially Burns’s answer on escalation. This is a public intelligence judgement, not disclosure of the underlying intelligence or proof of a Russian decision to use nuclear weapons.
9. Russian Federation, Presidential Decree No. 991, 19 November 2024, Fundamentals of the State Policy of the Russian Federation in the Area of Nuclear Deterrence. Russian text and English translation published by the US Naval War College’s Russia Maritime Studies Institute; particularly paragraphs 11, 15(g), and 18–20.
10. Michael Kofman, Anya Fink, and Jeffrey Edmonds, Russian Strategy for Escalation Management: Evolution of Key Concepts, CNA, April 2020. Analysis of Russian military writings; these writings should not be equated with a binding operational plan.
11. CIA, William Burns, remarks and discussion at Texas A&M University, 12 April 2023. Supports the attributed assessment of Putin’s advisory environment and risk-taking.
12. US Department of State, Office of the Historian, “The Cuban Missile Crisis, October 1962”. Official historical account of the confrontation and settlement.
13. Svetlana Savranskaya, editor, “The Underwater Cuban Missile Crisis at 60”, National Security Archive, 3 October 2022. Includes documentary material and participant recollections; the article avoids treating contested details of the near-launch episode as certain.
14. William Burr and Jeffrey Kimball, “Nixon’s Nuclear Ploy: The Vietnam Negotiations and the Joint Chiefs of Staff Readiness Test, October 1969”, National Security Archive, 23 December 2002. Documentary collection and historical analysis of the alert, its intended audience, and its results.
15. Stimson Center, “Nuclear Doctrine, Declaratory Policy, and Escalation Control”, 2004. Examines conflicting Indian and Pakistani interpretations of the 2001–02 confrontation.
16. Carnegie International Nuclear Policy Conference, 23 March 2015, “A Conversation with Gen. Khalid Kidwai”. Primary transcript of Kidwai’s explanation of Nasr and perceived room for Indian conventional operations.
17. Ministry of Foreign Affairs of the People’s Republic of China, 27 November 2025, China’s Arms Control, Disarmament, and Nonproliferation in the New Era, section on nuclear policy. Cited as China’s declared position.
18. US Department of Defense, March 2022, “Fact Sheet: 2022 Nuclear Posture Review and Missile Defense Review”. Used as a dated example of American declaratory policy, not as a claim that every subsequent policy formulation is identical.
19. Ministry of External Affairs, Government of India, Foreign Relations 2003. Official collection including explanations of the January 2003 nuclear doctrine, no first use, and the chemical and biological weapons reservation.
20. Jun Bong-geun, 21 November 2022, “Comparing North Korea’s Nuclear Forces Policy Laws”, Asia-Pacific Leadership Network. Analysis reproducing and comparing provisions of the 2013 and 2022 laws; used where an accessible original official publication was unavailable.
21. Robert Jervis, Perception and Misperception in International Politics, Princeton University Press, originally published 1976. Foundational scholarship on expectations, learning, cognitive bias, and decision-making under uncertainty.
22. Michal Smetana and Michal Onderco, “From Moscow With a Mushroom Cloud? Russian Public Attitudes to the Use of Nuclear Weapons in a Conflict With NATO”, Journal of Conflict Resolution, 67(2–3), pp. 183–209; first published online in 2022. Experimental findings do not establish subsequent trends in Russian opinion.
23. Astrid Kause, Helen Fischer, Zia Mian, and Susan T. Fiske, “People’s Responses to Nuclear Weapons: Mapping Post-Cold War Research”, Perspectives on Psychological Science, 2026. Review identifying limitations in the distribution and coverage of public-response research.
24. James M. Acton, “Escalation Through Entanglement: How the Vulnerability of Command-and-Control Systems Raises the Risks of an Inadvertent Nuclear War”, International Security, 2018; author’s summary and clarification published by Carnegie. See also his policy brief on nuclear command-and-control entanglement.