Bill Gates is seeking a meeting with Chinese President Xi Jinping to discuss a global approach to AI risks, including the possibility of monitoring whether models can help design harmful molecules or enable biological attacks. The proposal puts China at the center of a policy conversation that is often framed as a contest between Washington and Beijing rather than as a shared problem of controlling highly capable systems.
Reuters reported that Gates’ staff are working to finalize a China trip tentatively scheduled for November. He said he believes China might support restrictions on dangerous AI-model releases if the United States took the first step. China’s foreign ministry did not immediately respond to Reuters’ request for comment.
This is important to state plainly: it is Gates’ proposal, not an agreement with Beijing, and it is not evidence that Xi will meet him. The story is nevertheless noteworthy because Gates is arguing that AI safety requires cooperation with China, including on biosecurity risks that do not respect national borders.
A Proposal for Monitoring High-Risk AI Capabilities
Gates wants governments to monitor AI systems’ ability to design molecules and carry out biological attacks. He has called for an international organization informed by elements of nuclear inspections, aviation regulation, and agreements that protected the ozone layer. In a blog post, he argued for a global approach to managing a turbulent AI era.
The idea is not that every AI model should be banned or that AI research should stop. Gates’ argument is that certain capabilities, particularly those connected to biological harm, could require special oversight. The challenge is defining the threshold: a useful scientific model may assist legitimate research while also raising misuse concerns. Governments would need common tests, reporting mechanisms, and credible ways to respond when a model crosses a risk line.
China matters because it is a major AI developer and an increasingly important producer of open-weight models. A global monitoring system that excluded Chinese labs would be incomplete. A system that treated China only as a security threat would also be politically difficult to sustain. Gates is proposing that common risks could create room for cooperation even while the two countries compete over chips, models, cloud infrastructure, and export controls.
China Has Not Endorsed the Proposal
Gates told Reuters that China could be willing to limit releases of potentially dangerous models if the United States led by example. That is his assessment, not a public Chinese policy position. The foreign ministry had not responded when Reuters published its report, and no meeting with Xi has been confirmed.
The distinction matters because AI governance announcements are often overstated. China has promoted international AI governance initiatives and has developed domestic rules for services, data, and safety. But none of those facts amount to an endorsement of Gates’ proposed institution or a joint U.S.-China monitoring regime.
EastFrontier covered China’s recent statement that it has formulated nearly 200 AI standards, including ethics-review and service-practice work. The proposed global discussion would be different again from China’s AI companion rules, which illustrate how domestic regulation can target a particular category of user risk. Those domestic standards are relevant background, but they are not the same as an international mechanism to test models for biosecurity risk. Gates is calling for a separate layer of global coordination.
AI Biosecurity Could Test US-China Cooperation
AI safety often becomes entangled with commercial and geopolitical competition. Washington worries about Chinese access to advanced chips and model capabilities. Beijing criticizes U.S. technology restrictions and seeks a greater role in setting global AI rules. That climate makes it hard to build trust around data sharing, evaluations, and restrictions on model release.
Biosecurity may be one of the issues where the incentives are less divided. A model that materially lowers the barrier to creating a biological threat could affect every country, including the state that built it. The difficulty is turning that shared interest into practical standards without creating a system that one side sees as a tool of containment.
Gates is not the first person to argue for stronger AI oversight, but his decision to seek a conversation with Xi gives the proposal a specific diplomatic target. It also points to an unresolved question for China’s AI policy: can China combine rapid model development with a leading role in global risk management?
The coming months may provide an answer only if a meeting occurs and both sides choose to engage. Even then, a diplomatic discussion would be only the beginning. Officials would still have to decide which model capabilities merit testing, how companies report results, what safeguards apply to research access, and how any international body could operate without becoming another arena for strategic rivalry. For now, the facts are narrower. Gates has stated a proposal, his staff are pursuing a tentative November trip, and China has made no public commitment. That is enough to make the initiative worth watching, but not enough to call it a new international AI agreement.