This was made clear last week in La Coruña, Spain, when state delegations and representatives from the AI industry, academia and civil society convened the third Multi-Stakeholder Summit on Responsible Artificial Intelligence in the Military (REAIM). This summit aims to shape the future of international cooperation in this field. The past two summits produced an “outcome document” that was largely supported by the delegates in attendance. Both the 2023 “Call to Action” and the 2024 “Blueprint for Action” [PDF] Approximately 60 countries supported it. This year, only 35 countries, not including the United States or China, supported the outcome document, the Pathway to Action. [PDF].
Although non-binding, REAIM outcome documents typically include common-sense commitments, such as ensuring that militaries use AI in a manner consistent with international humanitarian law, and highlight what countries see as key concerns for the coming year. This decline in support for this year’s document is the latest example of ongoing macro-geopolitical divisions, particularly between the United States and Europe. The question now posed by REAIM is whether middle powers will promote AI rules of the road and confidence-building measures if the great powers become increasingly aloof.
The Donald Trump administration has disrupted U.S. relations, particularly with NATO partners. If countries are uncertain about how their own positions and relations with other countries will develop, with the United States and, to a lesser extent, China, it is difficult to commit to international cooperation or sign statements of principles that might be opposed by great powers. In fact, the number of US and Chinese delegations at REAIM in Spain was significantly smaller than at the 2024 summit in South Korea.
The growing gulf between accelerated AI integration efforts by militaries around the world is of concern to all countries, as international dialogue on military AI tends to highlight the risks and potential constraints to its use. Many of the traditional multilateral instruments for discussing the global governance of military applications of AI (including the United Nations Group of Governmental Experts on Lethal Autonomous Weapons Systems) continue at the glacial pace of international bureaucracy, as they have since the 2010s. But states are already developing and experimenting with AI capabilities, if not fully implementing, expanding, and deploying them. Ongoing conflicts such as Israel-Gaza and Russia-Ukraine are already using new AI tools, technologies, and response systems to create efficiency and power on the battlefield. As UN efforts to create binding regulations on military AI, particularly autonomous weapons systems, intensify, multilateral negotiations are at risk of becoming increasingly divorced from reality on the ground.
What the military now wants to do is figure out how to use AI safely and effectively, just as it has done with other technologies in the past. If this divergence continues unchecked, the risk doubles. In the long run, policy efforts can become divorced from the technical realities of the systems they seek to manage. In the short term, states are deploying these technologies in a patchwork of haphazard policies, without the opportunity to gain valuable insights into best practices from other countries.
Given the United States’ retreat from leadership in these areas, middle powers must now grapple with what to do and how to navigate confidence-building measures regarding military AI and cooperation. However, since the REAIM process has been led by middle powers from the beginning, this moment could be seen as an opportunity. The Netherlands launched this initiative in 2023. South Korea and Singapore hosted their second summit in 2024. And Spain hosted its third event last week. Two of the countries are NATO partners, and their relationship with the United States has fundamentally changed over the past year. As a result, they may feel uncertain about whether to continue their currently unpredictable partnership with the United States or pursue their own security goals more comprehensively.
One way forward is for mid-sized countries focused on AI adoption to drive the REAIM process. They were the first to create the REAIM process. They can use the momentum and convening power of the summit as a hub for international cooperation and capacity-building in military AI for non-superpower actors. And while the United States and China will always be invited, the middle powers need not worry about the extent of their participation. While this may make broad international agreement unlikely, especially if the summit process absorbs some of the capacity-building work previously done in the US-led Political Declaration on the Responsible Military Use of AI and Autonomy, REAIM could provide important capacity-building and rules of the road for middle powers and the Global South. The alternative would be to scale back initiatives such as REAIM and wait for Washington’s changing approach to the world to subside. This is wrong.
The REAIM process is an important bridge between diplomat-led UN efforts, which often focus on regulations and restrictions, and the realities of accelerating military investment and leveraging AI in a variety of use cases. As we saw in Spain last week, changes in the international landscape have made the role of bridge-building increasingly difficult, but it remains essential. Decisions made now could have ramifications through confidence-building measures and other opportunities to reduce the military risks of AI use, without restricting states’ use of this critical technology. If mid-power countries choose a more difficult path, they may be the ones to decide those outcomes.
This work represents only the views and opinions of the author. The Council on Foreign Relations is an independent, nonpartisan membership organization, think tank, and publisher that takes no institutional positions on policy issues.
