18 minutes
Summary
America’s word on nuclear matters is losing value: Allies increasingly doubt its promises, and adversaries increasingly dismiss its accusations. This credibility problem is often diagnosed as a shortfall in US military capabilities. It is better understood as a failure of signaling discipline: the ability to deploy intelligence disclosures in ways that reassure allies and deter adversaries. This article introduces “phantom signaling,” a growing problem in post-arms control nuclear statecraft. Phantom signaling is the invocation of intelligence to project resolve or reassurance without the evidentiary follow-through that makes a signal credible. It is “phantom” because the signal appears to be sent—the accusation is made, the guarantee offered—but nothing verifiable stands behind it. Intelligence is invoked but never substantiated. Deployed without defined thresholds (a stated line whose crossing triggers a response) or legible standards (criteria outsiders can see and apply themselves), intelligence becomes noise rather than signal, and disclosure becomes theater rather than statecraft: officials perform the act of signaling credibility without producing credible signals. Three recent cases show the same failure: the Intermediate-Range Nuclear Forces (INF) Treaty dispute, China’s low-yield testing controversy, and the Iran nuclear program. In each, phantom signaling actively erodes the credibility on which deterrence depends. Adversaries learn that accusations carry no defined consequences, and allies learn that guarantees rest on evidence they cannot evaluate.
Introduction
In a multipolar nuclear order, a state’s credibility—the belief that its threats and promises will be honored— is strategic capital. It shapes alliance cohesion and adversary risk calculations. Its durability can be spent to hold allies together, raise the risks that adversaries perceive in probing, and sustain norms that increasingly lack a treaty architecture to enforce them.
The United States has a growing nuclear credibility problem. Washington fields a large and sophisticated nuclear arsenal, maintains formal alliances with over 30 countries, and spends more on defense than the next several competitors combined. The problem is how effectively its threats and guarantees are believed—not necessarily the capabilities it has. What Washington increasingly lacks is signaling discipline across the full spectrum of nuclear statecraft: in its declaratory policy, in red-line articulation, in alliance assurances, and most acutely in the public invocation of intelligence. Deploying classified assessments in nuclear disputes is a particular kind of statecraft. It requires evidence that allies can anchor to, that adversaries cannot dismiss, and that compliance norms can absorb. Recent controversies over Russian treaty violations,[1] possible Chinese low-yield testing,[2] and Iran’s advancing nuclear program [3] show what happens when that discipline lapses. In each case, US officials cited classified assessments while offering limited public evidence, undefined compliance thresholds, and ambiguous consequences. Used this way, intelligence becomes noise rather than signal.
The architecture that once absorbed the credibility burden is thinning. As formal verification erodes, more of the work of maintaining nuclear stability falls on state signaling and reputation. In that environment, the difference between intelligence that signals and intelligence that merely generates noise is the difference between deterrence that holds and deterrence that drifts.
A Treaty-Thin World
For decades, formal agreements absorbed much of the credibility burden in nuclear politics—the work of making nuclear commitments believable—so that governments did not have to. Treaties defined thresholds of permissible behavior, established verification procedures, and provided institutional mechanisms to adjudicate disputes. They reassured allies that compliance would be monitored and communicated red lines to adversaries. That architecture is now significantly thinner. The INF Treaty is gone.[4] New START has lapsed.[5] The Comprehensive Nuclear-Test-Ban Treaty (CTBT) remains unratified. The Iran nuclear deal has collapsed.[6] As formal frameworks erode, that burden shifts to signaling—the actions by which governments communicate intentions, resolve, and reputation. Ultimately, that burden depends on predictability: whether governments have given adversaries and allies sufficient reason, through past conduct, to believe what they say and to predict what they will do.[7]
That shift creates a dual credibility problem. For adversaries, an accusation that names no threshold and carries no predictable consequence also provides no guidance about what behavior will actually be punished. Meanwhile, ambiguity about punishment invites probing (testing the line to see whether anything happens). For allies, credibility cannot function episodically; it must function continuously. Japan, South Korea, and NATO members should not have to reassess their security assumptions, weighing whether American security guarantees remain reliable enough to forego independent nuclear alternatives, every time Washington asserts, on the strength of intelligence it declines to share, that an adversary has crossed the line—an accusation of threshold-crossing that names neither the evidentiary standard it meets nor the consequence that follows. Allies are left to infer both what Washington believes and what it intends to do. The February 2026 China episode, discussed below, is the paradigmatic case of phantom signaling where accusations relied upon classified intelligence that was never revealed. In such a case, when a single signal must simultaneously constrain adversaries and reassure allies, its coherence is a strategic requirement, not a communications preference.
Extended deterrence raises the stakes further because it is where credibility is tested most directly. Unlike homeland deterrence, which rests on capabilities an adversary can count, extended deterrence depends on third-party perception: allies must believe not only that the United States possesses capability, but that it will interpret adversary behavior consistently and act predictably on its commitments. If compliance standards appear fluid or evidence appears selectively politicized, assurance weakens over time, through accumulated uncertainty that manifests in hedging strategies, defense diversification, and political hesitation. Debates over indigenous nuclear options that were unmentionable a decade ago are now openly conducted in serious policy circles in Seoul, Tokyo, and European capitals. Credibility erodes gradually, then irreversibly.
The Discipline Is Harder Now
Nuclear statecraft has always required calibrating disclosures of sensitive information. During the Cuban Missile Crisis in 1962, President John F. Kennedy declassified U-2 imagery of Soviet medium-range ballistic missile sites for Adlai Stevenson’s United Nations address, [8] while withholding the signals intelligence and human source reporting that had shaped the administration’s threat assessment. Two decades later, the Euromissiles Crisis [9] demanded a different kind of calibration. NATO governments needed sufficient detail on Soviet SS-20 deployments to sustain domestic support for the 1983 Pershing II and GLCM counter-deployments, but disclosure had to stop short of compromising the collection methods that made continued monitoring possible. The demand for calibration is unchanged; the structural environment in which it must occur is not.
Three structural shifts have widened the gap between secrecy and credibility. First, nuclear signaling—the declarations, deployments, alert changes, and disclosures by which a state communicates what it would do with its nuclear forces—is no longer directed at a single audience. In a multipolar system, a message intended for one audience may unsettle another, forcing Washington to reassure allies, deter rivals, and manage escalation across several fronts at once. Second, the evidentiary bar has risen. Iraq’s nonexistent weapons of mass destruction, mischaracterizations of progress in Afghanistan, and a series of contested intelligence claims since have left allies and adversaries discounting American assertions—even doubting them by default—in ways earlier generations of officials did not face: a discount earned outside the nuclear realm, now applied within it. Audiences that once deferred to official assurance now demand a demonstrable basis for a claim. Third, technological gray zones have widened. Low-yield testing, dual-use systems, and hypersonic delivery platforms blur compliance thresholds in ways that resist clear public explanation, and the burden of proof falls precisely on the kinds of classified assessments that are hardest to substantiate publicly.
Disciplined ambiguity—knowing what to reveal and what to withhold, and why—was never easy. [10] The cost of getting it wrong is now higher.
When Intelligence Stops Signaling
In principle, intelligence disclosure is a tool of statecraft. It can establish behavioral standards, anchor compliance judgments, and demonstrate to allies that American security guarantees rest on something more than assertion. [11] Three recent examples illustrate how this mechanism is now falling short. Each case exemplifies a distinct mode of phantom signaling: a sequencing failure in the INF Treaty dispute, a threshold failure in the China testing controversy, and a failure of message coherence in the Iran case.
INF Treaty
The INF Treaty dispute illustrates the sequencing failure, in which accusation, consequence, and evidence arrived in the wrong order. Beginning with the July 2014 State Department Compliance Report, the Obama administration publicly accused Russia of violating the INF Treaty—which banned both powers’ ground-launched missiles with ranges of 500 to 5,500 kilometers—by developing a prohibited ground-launched cruise missile. The unclassified report did not name the system. Allies received private briefings but no material they could use to build public consensus against Russia’s violation. The United States did not publicly identify the specific missile, the 9M729 (designated SSC-8 by NATO), until December 2017. The most detailed technical account of the violation did not come until Director of National Intelligence Dan Coats’ November 2018 disclosure, weeks after the Trump administration had already announced its intent to withdraw from the treaty. The sequence inverted the logic of evidentiary signaling: accusation came first, the decision to withdraw came second. Substantive public substantiation came third, deployed to justify a choice already made.A disciplined sequence would have run the other way: substantiation released alongside the accusation, allied consensus built on a shared record, and withdrawal announced as the stated consequence of continued violation. By the time the United States completed its withdrawal in August 2019, no shared evidentiary record had ever been established. The signal of accusation had been sent, but the substance never had to follow. Russia’s violation was real, but even a true accusation fails as a signal when substantiation arrives as justification rather than foundation.
China’s Low Yield Tests
The low-yield testing controversy illustrates the threshold failure: an accusation sustained for years without ever specifying what evidence, or what level of yield, would constitute a violation. Beginning with the 2020 State Department compliance report, US officials raised concerns about activity at China’s Lop Nur test site, including year-round operational preparations, explosive containment chambers, and blocking of data from the International Monitoring System (IMS), the global sensor network operated by the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO). All were framed as questions about Chinese adherence to the “zero-yield” standard for nuclear testing. [12] The concerns may be well-founded, but the public record has consisted only of suggestive circumstantial signatures and classified assessments that were never substantiated openly. In February 2026, US Undersecretary Thomas DiNanno used a Conference on Disarmament address in Geneva to accuse China explicitly of a yield-producing test on June 22, 2020, allegedly concealed through decoupling techniques.
Within 24 hours, CTBTO Executive Secretary Robert Floyd responded that the IMS had detected no event consistent with a nuclear test on that date. [13] The US accusation engaged only selectively with the IMS, the verification architecture that exists precisely to provide a shared evidentiary baseline, and the institutional reply made the gap visible. Without sustained engagement with that system, either to demonstrate its findings or to explain its limitations, Washington leaves the threshold question open: What yield, what instrumentation, what pattern of activity would constitute a violation? China can contest the basis indefinitely because no publicly legible standard has been articulated. The result is not productive ambiguity. It is a credibility vacuum that serves neither deterrence nor nonproliferation, and that weakens the very monitoring institutions whose authority matters most in a world without comprehensive test ban enforcement.
Iran’s Nuclear Program: From Imminent to Obliterated
The Iran case illustrates a failure of message coherence: official claims contradicted the intelligence and one another, as rhetoric substituted for evidence and for engagement with verification institutions. Following the Trump administration’s withdrawal from the Joint Comprehensive Plan of Action (JCPOA or Iran Deal) in May 2018 and Iran’s subsequent stepwise departure from its commitments—abandoning enrichment caps, restarting advanced centrifuges, breaching the 3.67 percent enrichment ceiling, and restricting International Atomic Energy Agency (IAEA) inspector access—US officials began citing shrinking breakout timelines with increasing frequency. At no point did they specify what Iranian action would trigger an American response. [14] The assessments rested on classified intelligence not shared with allies or Congress in actionable form. European partners, unable to evaluate the analytic basis, did not converge on the US position. [15] Public messaging compounded the problem by compressing Iran’s capabilities and intentions into a single claim, implying that enrichment capacity equaled weapons intent, a conflation the intelligence community itself resisted. In open testimony, Director of National Intelligence and CIA assessments consistently emphasized that Iran had not made a decision to pursue a weapon. Political messaging implied otherwise. [16]
The pattern reached its sharpest expression in June 2025, when US strikes on Iranian nuclear facilities were followed by a presidential declaration that the program had been “completely and totally obliterated.”[17] Within days, a leaked Defense Intelligence Agency assessment concluded that the facilities had been damaged but not destroyed, and that the program had been delayed by months rather than eliminated. [18] The gap between claim and intelligence was no longer inferential. It had become documented. What followed compounded it. The administration described Iran as actively rebuilding its nuclear program, then as posing an imminent nuclear threat, then as on the verge of collapse, each assertion made without a shared evidentiary basis, and each contradicting the last. [19] With IAEA access degraded by the strikes and congressional briefings reported as inadequate, the verification environment that might have stabilized interpretation had itself become a casualty. Phantom signaling normally means asserting what the evidence does not support. In this extreme case, it meant asserting what the evidence directly contradicted. Washington was asserting outcomes it could not demonstrate, in an environment it had helped make unverifiable. The consequence extends beyond this episode: Future American claims about nuclear programs—Iran’s or anyone else’s—will now be evaluated against a documented record of assertion outrunning intelligence, in an environment where the institutions capable of independent verification have themselves been compromised.
What Disciplined Ambiguity Requires
Across all three cases, the lesson emerges: phantom signaling does not merely fail to deter, it actively degrades the credibility on which deterrence depends. Adversaries learn that accusations carry no defined consequences, while allies learn that assurances rest on evidence they cannot evaluate. What disciplined signaling is meant to deter spans the cases: treaty violations, covert testing, unchecked proliferation—collectively, conduct that unravels what remains of the nonproliferation regime.
Two consequences follow, one behavioral and one institutional. Behaviorally, allies build institutional substitutes for credibility they can no longer take on assertion alone. The 2023 Washington Declaration, in which South Korea extracted a formal Nuclear Consultative Group as the price for forgoing independent nuclear options, is the clearest recent instance. What Seoul obtained was, at its core, access: standing consultation on nuclear planning and information-sharing that it could participate in rather than assurances it had to take on faith.
Institutionally, the organizations that stabilize nuclear politics lose the shared evidentiary baseline on which their authority rests. Iran is the sharpest expression. The June 2025 strikes terminated IAEA access precisely when independent verification would have mattered most: inspectors withdrew within weeks, and Tehran formally suspended cooperation in July. A year later, the agency reported it could not provide information on the size, composition, or whereabouts of Iran’s enriched uranium stockpile. [20] Assessments that followed flowed into that vacuum. Within a month, judgments of the damage derived from overlapping evidence emerged, ranging from a setback of months to many years. Whether the stockpile had been moved before the strikes was disputed as a matter of basic fact with a range of estimates that was not merely wide but unadjudicable. [21] The institution that would normally have narrowed the gap had been taken out of the game by the action under assessment. The result is an environment in which no party—ally, adversary, or inspectorate—can establish what is true about the program at issue. The deeper cost falls on the next dispute, which will arrive with no way to settle it.
The solution, however, is not maximal transparency. Strategic stability has long rested on uncertainty—on the mutually held recognition that some questions about capabilities, intentions, and thresholds are more stabilizing when left unanswered rather than resolved. Rigid publicly declared thresholds narrow escalation options and can lock states into commitment traps in crises—positions they must escalate to defend; [22] detailed disclosures can foreclose the diplomatic latitude that allows the adversary to back down without humiliation; revealing sources and methods can compromise the intelligence collection process upon which future judgments depend. Each of these costs is real, as the cases above confirm. The task is not to eliminate ambiguity but to discipline it—to preserve uncertainty where it stabilizes deterrence and reduce it where miscalculation would be catastrophic, where allied confidence cannot be sustained on assertion alone, or where adversary behavior must be held to a publicly legible standard. That discrimination, between the ambiguity that stabilizes and the ambiguity that corrodes, is what disciplined ambiguity demands and what recent practice has ceased to provide.
The US intelligence declassification campaign preceding Russia’s 2022 invasion of Ukraine shows what disciplined ambiguity looks like in practice. Washington selectively released satellite imagery, intelligence assessments, and warnings about Russian false-flag preparations without exposing sources wholesale or publishing full intelligence dossiers. [23] It revealed enough to shape international perception, preempt disinformation narratives, and strengthen allied cohesion. Rather than deterring the invasion, it worked to strategic advantage: Russia’s narrative space was constrained, partners were unified, and American credibility was enhanced rather than spent.
What distinguished this episode was its structure. Disclosure was tied to specific behaviors and bounded by specific purposes. Where the INF Treaty dispute put accusation before evidence, the Ukraine campaign released evidence before the consequence it warned of. Where the China accusations left the threshold undefined, the Ukraine warnings named the line in advance. And where the Iran case substituted rhetoric for engagement with verification frameworks, the Ukraine campaign worked alongside open-source analysis and allied intelligence services rather than around them. The threshold was named in advance: National Security Advisor Jake Sullivan warned publicly that any military action preceded by a staged pretext would be identified as Russian aggression before Russia could deploy it. The evidence was public: satellite imagery documented troop positions at specific locations along the Ukrainian border. And the demanded behavior was clear: stand down, or be seen by every ally and partner as having crossed a line Washington had defined in advance. [24]
That structure is what the INF, China, and Iran cases lack. Restoring it means treating intelligence disclosure as an instrument of statecraft rather than episodic messaging—and that, in turn, requires four disciplines that recent practice has neglected. First, evidentiary thresholds should be defined before public accusations are made because precommitment reduces the perception of ad hoc politicization and gives both adversaries and allies a standard against which to evaluate any claim. Second, disclosure should be tied to a strategic purpose. Deterrence and alliance reassurance each require different levels of transparency and should not be collapsed into a single institutional reflex. Third, where verification mechanisms exist, they should be visibly engaged; even partial public alignment with established monitoring bodies raises the evidentiary bar for adversarial contestation and reinforces institutional legitimacy from which the United States ultimately benefits. Fourth, intelligence rhetoric should be deployed selectively. Routine invocation of “we have intelligence” trains adversaries and allies alike to discount the signal—and exhausts the credibility reservoir that disciplined disclosure is meant to protect.
None of these is a radical proposal. Each represents a return to the practices the United States has demonstrated it can sustain when disclosure is treated as a strategic instrument rather than a rhetorical reflex. [25]
Conclusion
The INF Treaty, China, and Iran cases point to the same conclusion: phantom signaling actively spends the credibility on which deterrence depends. Adversaries learn that American accusations carry no defined consequences. Allies learn that guarantees rest on evidence they cannot evaluate. Rebuilding that credibility requires coherence: alignment between evidence, thresholds, and consequences—legible to adversaries, anchorable by allies, and absorbable by the residual architecture of arms control norms. No amount of assertion—invoking intelligence, resolve, or red lines—closes that gap.
The 2022 declassification campaign showed that this credibility is achievable. As formal arms control constraints weaken, the burden on signaling will only increase. The challenge ahead is to restore the discipline—structuring uncertainty where it stabilizes deterrence, reducing it where miscalculation would be catastrophic, and ensuring that the signals Washington sends are legible to the audiences that matter most. In a world where the institutional architecture that once carried much of this burden has eroded, signaling discipline is not a communications preference, but a condition of strategic stability.
