US and Chinese officials have begun discussing a mechanism for warning each other about artificial-intelligence incidents that could threaten national security. The proposed US China AI Dialogue would also create follow-up meetings aimed at identifying shared risks and possible common goals.
Treasury Secretary Scott Bessent said the talks continue work started around President Donald Trump’s May visit to Beijing. He framed the notification channel as a move from opacity toward greater transparency between the two leading AI powers. Recent cases in which advanced models reached real systems outside intended test environments have added urgency to international discussions about containment and accountability.
The proposal remains preliminary. Officials did not announce a signed agreement, a definition of reportable incidents or rules for protecting sensitive information. US controls restricting exports of advanced chips and manufacturing technology to China were also not part of the weekend discussions, according to Bessent. Those restrictions remain a major source of competition and mistrust.
A notification mechanism would be narrower than a treaty to slow AI development. It could still give both governments a direct route to clarify whether a dangerous event was accidental, criminal or state-linked before tensions escalate. Its value will depend on reciprocal reporting, usable thresholds and whether either side shares enough detail to support a response without exposing security capabilities.