For years, artificial intelligence has been the elephant in the room every time American and Chinese officials sat across from each other. They talked trade. They talked Taiwan. They talked tariffs and rare earths and semiconductor export controls. But AI — the technology both nations have publicly identified as the defining strategic competition of the 21st century — has been largely conspicuous by its absence from any formal diplomatic channel.
That is about to change.
With President Donald Trump heading to Beijing for a summit with President Xi Jinping on May 14–15, senior officials on both sides are now actively working to put AI governance on the agenda for the first time under the Trump administration. If it happens, the conversation will mark a watershed moment: two superpowers locked in an increasingly fierce technological arms race sitting down to discuss, at least in some limited way, the risks that race poses for everyone.
Why Now?
The timing is not an accident. The past six months have been extraordinary in the history of AI development. Anthropic’s revenue surged 80x year-over-year. Four major Chinese labs — DeepSeek, Kimi, MiniMax, and Zhipu AI’s GLM — released top-tier open-weight models within a 12-day window, with none costing more than a fraction of comparable Western offerings. A Harvard Medical School study published in Science found that OpenAI’s reasoning model outperformed experienced ER physicians in diagnostic accuracy. And intelligence agencies from six allied nations issued an unprecedented joint warning that autonomous AI agents are being deployed without adequate safety governance.
In short: the technology is moving faster than the policy frameworks designed to contain its risks. Both Washington and Beijing know it. And as the Trump-Xi summit approaches, there is a growing recognition in both capitals that some degree of dialogue — however limited — is better than none at all.
“The United States and China hardly talk about AI, at least at the official level,” one analyst with knowledge of the preliminary discussions told reporters this week. “Both nations see it as a key strategic technology. But their relationship in this domain is characterized by a low degree of trust and a high degree of competition. What’s being proposed isn’t a partnership. It’s the beginning of a conversation.”
What’s on the Table
According to multiple reports, the proposed AI dialogue would focus on what officials are calling “frontier risk” — the dangers posed by unpredictable AI models, autonomous weapons systems, and the potential for non-state actors to exploit advanced open-source technologies. The talks would not touch the core competitive dimensions of the AI race: chip exports, model access restrictions, or military AI development programs.
On the American side, Treasury Secretary Scott Bessent is reportedly leading the early groundwork for the AI track at the summit. China has yet to formally designate its counterpart, though Vice Finance Minister Liao Min has been involved in preliminary discussions.
Analysts are careful to keep expectations calibrated. “We should not expect a landmark agreement or a Geneva Convention for AI,” said one foreign policy researcher cited by multiple outlets. “What we might get is a structured mechanism for ongoing communication — a hotline, essentially, for AI-related risks. That would itself be significant.”
The Stakes Are Real
The urgency behind these talks is not theoretical. The convergence of AI capabilities with military applications is accelerating at a pace that gives even hawks on both sides pause. The U.S. Air Force’s WarMatrix AI-powered wargaming environment just completed its first operational exercise in March. China has its own AI-enhanced military simulation programs. Autonomous systems capable of making lethal decisions in milliseconds are no longer science fiction — they are procurement items.
At the same time, the rapid proliferation of open-weight models from Chinese labs has scrambled the assumptions underlying U.S. export controls. If a state or non-state actor can download a frontier-level AI model with no restrictions, the carefully constructed semiconductor export regime begins to look like a fence around a hole in the ground.
The Trump administration has been notably quiet on AI safety governance compared to its predecessor, which signed an executive order on AI safety in 2023. But the geopolitical logic of at least establishing communication channels with China appears to be cutting through internal ideological resistance.
Geopolitics of the Algorithm
The broader context of the Trump-Xi summit is itself significant. The two leaders are meeting against a backdrop of ongoing trade tensions, renewed discussions over Taiwan, and a complex shared interest in managing the fallout from the Iran situation. AI is entering this diplomatic space not as an abstract technological concern but as a live operational issue intersecting military readiness, economic competitiveness, and strategic risk.
There is, of course, deep skepticism in both countries about whether talks will lead anywhere meaningful. Critics in Washington argue that any AI dialogue gives China intelligence about American concerns and red lines. Critics in Beijing view American AI governance initiatives as thinly veiled attempts to constrain Chinese technological development under a safety framework designed by the West.
Both critiques have merit. And yet the alternative — two nuclear-armed superpowers developing AI-enhanced military systems with zero communication about the rules of engagement — arguably carries greater risks.
A Moment That Won’t Come Twice
Whether the Trump-Xi summit produces an AI dialogue framework or simply a vague communiqué promising future talks, the willingness to engage at the diplomatic level represents a shift. For the past several years, the AI competition between the United States and China has been conducted almost entirely through market forces, research publications, export controls, and — increasingly — open-source model releases that scramble any one actor’s attempt to dominate the field.
Diplomacy is catching up. Slowly, haltingly, with deep mutual suspicion on both sides. But catching up nonetheless.
The summit begins in six days. The world’s most consequential technological rivalry may, for the first time, acquire a diplomatic dimension. Whether that dimension will matter — whether it can produce the kind of minimal trust that prevents catastrophic accidents — is a question that won’t be answered in Beijing next week. But it might, at long last, begin to be asked.
Sources: CommonWealth Magazine, Benzinga, Invezz, Voice of Emirates, MarketDash, IBTimes Singapore, EconoTimes, HNGN, CNBC
