In February 2026, more than 200 personnel from Australia, the United Kingdom, and the United States gathered on Australia’s east coast for the latest AUKUS Maritime Big Play exercise, where operators tested roughly thirty next-generation autonomous capabilities, including a live strike by a one-way attack munition. Earlier iterations of the series had already demonstrated something more consequential than any single platform: the ability of the three navies to control autonomous assets belonging to one another. Allied autonomy has moved out of the laboratory and into operational formations, and it is moving faster than the alliance frameworks that govern who may authorize a machine to act. 

The Seam in the Umbrella 

Extended deterrence has always rested on shared expectations about decision-making. The United States and Japan formalized this in December 2024, when the two governments issued their first Guidelines for Extended Deterrence, reinforcing consultation and communication procedures for the gravest decisions the alliance could face. Consultation procedures of this kind assume that the decisions in question move at human speed, through ministers and commanders who can meet, confer, and align. Autonomous systems introduce a class of alliance decisions that will not wait for a meeting. 

Every ally that fields autonomous systems encodes national rules about when a machine may engage, when it must ask a human, and when it must stop. The United States codifies its approach in Department of Defense Directive 3000.09, which sets rigorous verification, testing, and senior-review requirements designed to minimize unintended engagements. Allies maintain their own rules, shaped by their own laws, doctrines, and political constraints. 

Those national rule sets do not automatically align, and in a combined force the differences become seams. A coalition formation may contain systems that would engage a given contact under one national authority profile and hold fire under another. Adversaries study seams. 

The probing will not be hypothetical. China has spent more than a decade practicing pressure beneath the threshold of armed conflict, from maritime militia swarms to coercive air intercepts, precisely because ambiguity about an alliance response is exploitable. Russia has run the same playbook against NATO members in the air and undersea domains. A coalition whose autonomous systems answer that pressure under visibly different national rules hands its adversaries a map of where the alliance hesitates. 

The consequences for deterrence are direct. The most immediate risk is entrapment at machine speed: when an allied autonomous system engages inside the American security umbrella, the United States can inherit an escalation decision it never made, at a pace that forecloses the consultation the alliance promised itself. Mismatched authority thresholds also invite probing, because an adversary can calibrate provocations that one member’s systems would answer and another member’s system would not, and can harvest knowledge of alliance rules from the pattern of responses. 

Assurance suffers as well. An alliance that cannot state clearly how its machines hold fire loses precision in its guarantees at exactly the moment allies under threat are asking for more precision, not less. Credibility in the machine age includes the credibility of restraint. 

Authority Interoperability 

The building blocks for closing the seam already exist. Directive 3000.09 gives the United States a mature national template. The Five Eyes cybersecurity agencies showed in May 2026 that allied governments can speak with one voice on machine autonomy when they jointly published guidance on agentic artificial intelligence, urging organizations to prioritize “resilience, reversibility and risk containment over efficiency gains.” The Maritime Big Play program has likewise turned toward shared command-and-control software and a common autonomous baseline across the three AUKUS navies. Each of these efforts is necessary. What none of them yet delivers is the specific artifact extended deterrence requires: a legible statement of each ally’s engagement authority rules that coalition systems and coalition commanders can rely on before they operate together. 

Allied autonomy needs authority interoperability, built as deliberately as data interoperability was built over the past three decades. Three practical steps would supplement the work already underway. Allies could publish national authority profiles, meaning standardized declarations of the engagement tiers their systems may exercise, the conditions that trigger human review, and the recall mechanisms that stop a pending action. Combined forces could adopt shared reversion standards, so that any coalition system can be halted through a common, tested procedure regardless of which flag it flies. Extended deterrence consultations, on the model the United States and Japan set in 2024, could add an autonomy annex that establishes in advance how allied autonomous engagements inside the umbrella will be communicated, reviewed, and, where necessary, walked back. 

These steps ask allies to formalize what alliance staffs already negotiate informally under time pressure. Deterrence theory has long held that credibility rests on capability and will. Combined autonomy adds a third term: legibility. An alliance whose machines carry legible, compatible, and reversible authority is an alliance an adversary cannot divide at the seam. Extended deterrence in the machine age will belong to the alliances that make their authority architecture as interoperable as their ammunition. 

Burak Oktenli is an independent researcher in Washington, DC, conducting research on human-machine authority architecture and the governance of autonomous military systems. He is the author of the ten-volume Authority and Architecture series on the governance of autonomous decision authority, and his analysis has appeared in RealClearDefense, Task and Purpose, and The Space Review. Views expressed in this article are the author’s own. 

Allied Autonomy Is Creating a New Seam in Extended Deterrence was originally published on Global Security Review.

Historically, adversaries, ranging from Imperial Japan to Saddam Hussein’s Iraq, badly miscalculated American resolve. This was, in part, because the US did not effectively communicate national will and the country’s ability to generate combat power.

While nuclear deterrence held during the Cold War, a worrisome erosion of America’s nuclear deterrence appears underway. A combination of unilateral cuts to the American arsenal, sluggish nuclear modernization, and “nuclear psychosis” (fear of one’s own nuclear arsenal) likely led Russian President Vladimir Putin to develop an escalate to win nuclear doctrine. The same factors may also contribute to China, Iran, North Korea, and Russia issuing mounting conventional and nuclear threats against the West.

Americans need to remember that escalation dominance is an essential ingredient in effectively maintaining deterrence of adversaries and assurance of allies. Escalation dominance means having the option to retaliate up or down the escalation ladder—at America’s choosing. This means developing and fielding capabilities that at least achieve parity, if not superiority, at each rung of the escalation ladder.

Without the US investing in the needed capabilities for full-spectrum conflict, all-domain escalation dominance, adversaries can study America’s capabilities and identify gaps. They can then develop and execute successful strategies that exploit those gaps. They can also incorrectly believe gaps exist, because of a failure of American messaging. Either way, American deterrence might fail.

The US once understood the art and science of escalation dominance—including how to weigh and manage uncertainties and unavoidable risks. At the most dangerous point in the Cold War, the Cuban Missile Crisis, the US out-thought, outmaneuvered, and overmatched the Soviet Union. Soviet Premier Nikita Khrushchev backed down. President Ronald Reagan successfully escalated an arms race while also waging economic warfare. The Soviet Union collapsed.

America now faces multiple adversaries it must engage across the spectrum of conflict. Both horizontal and vertical escalation are certain. With the potential for conflicts to start in the cyber or space domain, the United States must have the ability to limit damaging incursions and malicious interference, while out-escalating an adversary.

History shows that America’s enemies, whomever they might be, tend to be relentless, brutal, casualty acceptant, and even genocidal. Current events, in conflict zones around the world, prove that this has not changed. Appeasement is as ill-advised today as it was with Nazi Germany and Imperial Japan. Restraint is also as ill-advised now as it was then.

Sound theories of American victory need to be developed. Sound American strategy requires the proper matching of objectives to means. Without these in place, deterrence might not hold.

The stakes could not be higher. Conventional war is still almost always a prolonged and very bloody business, in which civilians are caught in the middle and have no safe refuge. If America’s strategic deterrence ever fails, the homeland might suffer devastating attack. Destruction in Ukraine is a small example of what such destruction could look like.

Nuclear war would likely prove unimaginably worse. Nuclear deterrence demands the ability to respond with proportional retaliatory strikes. It also requires the ability to match enemy escalation. The US cannot allow itself to be self-deterred by either a lack of will or a lack of capability. Self-deterrence is to give license to adversaries.

Yet some in the US continue to advocate for policies that eschew escalation at any cost.  This is a theory of defeat, not victory. Attempts to avoid risks only increase the cost of risk.

Such advocates need to be reminded that bullies will always test those they despise. Saving face and taking safe off-ramps are not a priority. For the US to be unable or unwilling to respond with sufficient force when necessary will only invite further and far worse aggression.

In the Cold War now underway, there is a danger of rapid escalation. To cede the military initiative to adversaries, to abandon America’s time-honored strategic culture of retaining full-spectrum escalation dominance, is to invite a nuclear strike against the nation and/or its allies. Adversary malign-influence campaigning is underway and now pervades American information.  Succumbing to these efforts to convince American society that isolation and pacifism are the nation’s best option is a mistake.

Appeasement never satisfies autocratic aggression. The United States must instead invest in the capabilities required to re-master the art and science of strong-stomached escalation dominance. This will support deterrence of America’s adversaries, assurance of America’s allies, and defense of the homeland. Now is the time to act.

Joe Buff is a Senior Fellow at the National Institute for Deterrence Studies. Views expressed in this article are the author’s own. 

Modern Escalation Dominance Is Essential to Effective Deterrence and Assurance was originally published on Global Security Review.