When the Referees Leave the Field: Military AI Is Advancing Without a Rulebook

Introduction

In February 2026, diplomats, defense officials, and technology leaders gathered in A Coruña, Spain, for the third international summit on Responsible Artificial Intelligence in the Military Domain — a process known as REAIM. These summits have been, since their inception in 2023, one of the few serious multilateral efforts to establish guardrails on how governments use artificial intelligence in warfare.

The outcome was sobering. Only 35 nations signed the summit's outcome document, "Pathways to Action" — roughly half the number who endorsed the equivalent document at the 2024 summit in Seoul. The United States and China, the two dominant powers in both AI development and military capability, were not among them.

This is not just a story about diplomatic failure. For those of us working at the intersection of military communication, public affairs, and AI ethics, it is a signal worth taking seriously.

What REAIM Is — and Why It Matters

REAIM is not a treaty body. It produces no binding regulations. What it produces are "outcome documents" — statements of common-sense principles, such as commitments to use AI in ways that comply with international humanitarian law. You might wonder: if the commitments are non-binding, why does the withdrawal of major powers matter so much?

Because these documents do something treaties can't do quickly: they shape norms. They tell militaries, defense contractors, and technology developers what the community of nations considers responsible. They create political costs for bad behavior and establish a shared vocabulary for what accountable military AI looks like. Their value is reputational and epistemic, not legal.

When 60 nations endorse a document in 2023, and 35 do in 2026, the norm-building machine is running in reverse.

The Communication Failure Nobody Is Naming

From a public affairs perspective, the story underneath this story is about strategic communication — or its absence.

The United States sent a smaller delegation to A Coruña than it sent to Seoul. China did too. That absence communicates something, whether intended or not. It signals that the current U.S. administration does not view multilateral AI governance as a priority. It signals that the world's two leading AI powers are, at least for now, opting out of the effort to define what responsible military AI looks like globally.

No press release announces this. No spokesperson explains it as policy. But the message is received. And in the absence of an authoritative, consistent U.S. voice in these spaces, other actors — adversaries, partners, and observers — will fill the interpretive vacuum.

This is a strategic communication problem of the first order. And it is exactly the kind of problem that military public affairs professionals and communications strategists need to understand, because navigating it requires more than messaging. It requires institutional clarity about values.

The Middle Powers Moment

The authors of the CFR analysis, Michael Horowitz and Lauren Kahn, make a compelling argument: the middle powers who built REAIM — the Netherlands, South Korea, Singapore, and Spain — should not wait for Washington or Beijing to re-engage. They should advance the process themselves.

This is not just diplomatically wise. It is strategically necessary. NATO partners are already uncertain about their relationships with the United States. Countries in the Global South are watching to see whether responsible AI governance is a Western export or a genuinely universal project. If middle powers can sustain the REAIM process, build capacity, and create practical rules of the road, they may produce something more durable than any great-power-dominated framework could achieve.

There is a lesson here for domestic institutions as well. When large, powerful actors disengage, smaller actors have a choice: wait for the powerful to return, or build something worth returning to.

What the Battlefield Is Already Teaching

The authors note that ongoing conflicts — Israel-Gaza, Russia-Ukraine — are functioning as live testing environments for military AI. Targeting assistance, autonomous drone systems, cyber operations, and AI-enabled intelligence analysis are not theoretical. They are active.

This matters for governance because the gap between field deployment and international regulation tends to widen, not narrow, over time. The more AI systems become embedded in military doctrine, the harder it becomes to regulate them after the fact. The window for establishing meaningful norms — even informal ones — is not indefinitely open.

For anyone engaged in military communication or public affairs, this also means that AI-generated content, AI-assisted decision-making, and AI-enabled information operations are increasingly part of the operational environment. Understanding these tools, and the ethical frameworks that should govern them, is no longer optional.

Practical Takeaway

If you are a military communicator, a public affairs officer, a policy professional, or a student of strategic communication, here is the practical implication of the REAIM story: the institutions that govern AI in warfare are under construction, and the construction site is often empty.

That creates risk. It also creates opportunity. Organizations and individuals who engage now — in academic forums, interagency working groups, professional military education, and public commentary — can contribute to the frameworks that will eventually govern these systems. The alternative is to arrive late and inherit whatever structure others built.

The referees didn't leave the field because the game got too complicated. They left because staying required leadership that nobody wanted to claim. That is both a diagnosis and an invitation.

Closing Reflection

I have spent years in military public affairs learning that institutional communication is never just about messages — it is about trust, credibility, and the long-term relationship between an organization and its publics. The same principle applies at the level of nations and multilateral institutions.

What REAIM 2026 revealed is not a technical problem. It is a leadership problem. And leadership problems, by definition, can be solved by people willing to lead.

The question is who steps forward.

---

Sources

Horowitz, Michael C. and Lauren Kahn. Military AI Adoption Is Outpacing Global Cooperation. Council on Foreign Relations. February 11, 2026. https://www.cfr.org/articles/military-ai-adoption-is-outpacing-global-cooperation

REAIM 2023 Summit. The Hague, Netherlands. February 2023. https://www.government.nl/topics/reaim

REAIM 2024 Summit. Seoul, South Korea. September 2024.

REAIM 2026 Summit. Pathways to Action. A Coruna, Spain. February 2026.