CNAS Insights | Trump Should Talk to Xi About Military AI

When President Donald Trump goes to China to meet with General Secretary Xi Jinping next month, the leaders of the world’s two superpowers will have much to discuss, with trade and the Middle East taking top billing. In this crowded agenda, the military use of artificial intelligence (AI) deserves attention as an issue where the two sides could agree to measures designed to prevent accidents and enhance general stability without putting unrealistic constraints on military-technological innovation.

In 2024, after a year of grinding diplomacy, Xi agreed with Trump’s predecessor, Joe Biden, to ensure humans, rather than AI, would always make the final decision on using nuclear weapons. When Trump and Xi previously met in South Korea last October, China raised AI as a general area for cooperation. But it was not discussed in depth, and no follow-on discussions have taken place. Treasury Secretary Scott Bessent now suggests the issue may be on the table in Beijing.

Trump has an opportunity to go much further, demonstrating American strength in the pursuit of a “decent peace,” as the 2026 National Defense Strategy put it, for the AI era. He could begin with nuclear weapons. The president has rightly called for arms control talks with China (and Russia) to uphold strategic stability—reaffirming a commitment to human control aligns with his goals.

The Trump administration could make breakthroughs in two new areas. One is moving discussions of AI in nuclear operations to the working level, where experts could hammer out what implementation looks like in concrete terms. Another opportunity is expanding AI talks to cover conventional weapons. Ongoing wars in the Middle East and Ukraine—which have seen AI used for planning and targeting strikes—only adds urgency to the need for discussions between the world’s two leading militaries. U.S. forces integrating AI into their campaigns against Venezuela and Iran demonstrate America’s advantage relative to a Chinese military whose AI aspirations are clear but untested.

As Washington and Beijing continue to operate militarily in East Asia and cyberspace, they need to explore where technological innovations require changing the rules of engagement. For example, what happens when an AI system launches a cyberattack against the wrong target, or when an autonomous drone ship from country A collides with one from country B?

Exchanging views can help the two powers distinguish between accidents and intended uses of the AI-enabled weapons they have now and will build more of as the technology matures. Over time, the two sides could move toward agreeing on norms and standards. Washington has a preexisting set of principles for the ethical use of AI in the military, reflected in an international framework launched in 2023 that 54 other countries now support. The two leaders could refresh these principles and order newly revived military-to-military talks to flesh them out.

Critics will say that diplomatic processes encourage accommodation to Beijing. That is a misunderstanding. Diplomacy done right provides a channel to convey firm messages. It can bolster deterrence rather than invite aggression, informing rivals about how the United States intends to respond to specific behaviors.

Skeptics will also warn that any potential agreement would constrain the United States more than China. Beijing’s poor record complying with previous bilateral commitments, such as the 2015 deal on cybersecurity, is cause for vigilance. But the president should be confident in the U.S. position. Having an established policy on the military use of AI means American diplomats can communicate in a clear, authoritative manner to promote responsible behavior rather than constrain U.S. capabilities.

The United States and China will remain fierce rivals competing across every dimension of global politics, and AI is an opportunity to gain an edge. Washington and Beijing will not resolve their disagreements or cooperate jointly. There are undoubted limits on what is possible. At best, talks can help prevent AI from precipitating crises and enable leaders to arrest a spiral into conflict—no small feat.

In fact, military AI remains the most tractable issue in what could be a sprawling issue set. Washington and Beijing share a common interest in strategic stability and avoiding unintended escalation. Few other topics are so aligned. Mutual distrust about COVID-19 and Beijing’s use of AI in its global dragnet of dissidents complicates discussions on the misuse of AI. American and Chinese companies compete fiercely to innovate and win global market share. Ideas about how to deal with societal dislocation diverge as much as each country’s political systems.

Finally, while the United States and China will sit at the negotiating table, the entire world watches. Washington can improve its global reputation through good-faith engagement with Beijing on military AI. If talks work, the risks of superpower conflict that would affect other countries go down. If China withdraws or grandstands, Washington looks responsible while Beijing looks recalcitrant.

AI’s importance to modern warfare grows with every new conflagration. By talking about military AI, the United States can signal strength toward China and simultaneously prop up peace in a moment when it faces many tests. Trump has an opportunity to define the rules of the road for a technology that will shape conflict for a generation.

Jacob Stokes is a senior fellow and deputy director of the Indo-Pacific Security Program at CNAS where his work focuses on U.S.-China relations, Chinese foreign and military policy, East Asian security affairs, and great power competition.

Daniel Remler is a senior fellow with the Technology and National Security Program at CNAS. His research focuses on the implications of AI and emerging technologies for U.S. national security, foreign policy, and strategic competition, including risks from advanced AI systems.