U.S. Proposes AI-Incident Alerts With China—But No System Exists Yet

Verified Baseline

The United States proposed a notification mechanism

After talks in New York with Chinese Vice Premier He Lifeng, Treasury Secretary Scott Bessent said the United States had proposed a mechanism through which the two governments could notify one another about artificial-intelligence incidents that rise to a national-security level. The proposal was reported independently by the Associated Press and Reuters.

The two sides also discussed establishing a broader U.S.–China AI dialogue ahead of a planned meeting between President Donald Trump and Chinese President Xi Jinping. Chinese state media described the talks as candid and constructive but did not publicly detail a notification system.

What is confirmed: the U.S. made a proposal and both sides discussed AI risks. What is not confirmed: that China accepted a final system, what events would trigger notice, how quickly notice would be required or whether disclosures could be independently verified.

What It Could Cover

National-security incidents remain undefined

Potentially relevant events could include AI-enabled cyberattacks, dangerous model behavior, biological misuse or a loss-of-control event. Those examples have been raised by specialists in the broader policy debate, but officials have not published a negotiated list. Treating any particular category as part of the agreement would therefore go beyond the available evidence.

A useful notification system would need at least four elements: a shared threshold for reporting, a secure communication channel, rules protecting sensitive information and a process for confirming that an alert is credible. None has yet been released publicly.

What It Does Not Mean

This is not an AI treaty, a pause or a joint regulator

The proposal does not slow model development, change export controls or create a supranational agency. It also does not resolve the competition between the two countries over advanced chips, open-source models, military applications and commercial leadership.

Incident notification is closer to crisis communication than shared regulation. Its value would depend on whether each government trusts the other enough to report quickly and candidly during a serious event—precisely when strategic incentives may favor secrecy.

Why It Matters

A narrow safety channel could reduce misunderstanding

The United States and China host many of the organizations developing advanced AI systems. A reliable alert channel could help officials distinguish an accident, criminal act or uncontrolled technical failure from deliberate state action. That distinction could matter during a fast-moving cyber or security crisis.

The proposal’s significance should not be overstated. A communications channel cannot prevent an incident by itself, and it may fail if the parties disagree about what counts as national security or fear that disclosure would expose capabilities and vulnerabilities.

The Next Evidence

Watch the leaders’ meeting and any written terms

The assessment would change materially if the White House or Chinese government publishes a joint statement, definition of covered incidents, implementation timetable or named points of contact. Evidence that the mechanism has been tested or used would be stronger than a general commitment to continue dialogue.

Until then, the accurate description is limited: Washington proposed an AI-incident notification mechanism, Beijing discussed AI issues, and the operational system remains unresolved.