AI talks move from agenda to a proposed notification mechanism
US Treasury Secretary Scott Bessent said after talks in New York with Chinese Vice Premier He Lifeng that Washington proposed a notification mechanism for artificial-intelligence incidents that could affect national security. Associated Press reports the proposal as part of the bilateral discussions, while Reuters says the two sides agreed to continue a formal dialogue on AI safety and describes an incident line intended to improve communication during serious AI events.
China’s State Council, carrying a Xinhua readout, independently confirms that the delegations held dialogue on AI-related issues during the consultations. The Chinese account does not say that Beijing accepted an incident-notification channel and does not publish operating terms for one. That difference matters when separating the confirmed talks and US proposal from the more detailed bilateral mechanism described by Bessent through Reuters.
The proposed channel has no published operating rules yet
Bessent framed the mechanism as a way for the two largest AI powers to improve transparency and develop a shared understanding of risks that could affect both countries. Public descriptions point to notification and communication during significant AI safety or national-security incidents rather than a broader agreement on model development, export controls or deployment policy.
Neither government has published thresholds for triggering a notification, which agencies would operate the channel, what information would have to be shared or whether participation would be binding. The available evidence therefore does not support describing the proposal as a treaty, a permanent crisis hotline or a common technical safety standard.
Bessent says the next AI meeting is planned for Shenzhen
Reuters reports that Bessent expects senior US and Chinese officials to meet again in roughly two months in Shenzhen to continue the AI-safety discussion. That gives the dialogue a prospective follow-up beyond the New York consultations and moves the subject further than the pre-meeting agenda described in the earlier version of this story.
The Shenzhen timing and location are Bessent’s account. China’s published readout confirms the AI dialogue but does not announce that follow-up meeting or identify participants. The New York consultations also covered trade and other economic issues, so the AI track remains one part of a wider bilateral relationship with major disagreements still intact.
What changed and what remains unsettled
The material change is that US-China AI risk discussions have now taken place at senior government level, Washington has put a concrete incident-notification mechanism on the table, and Reuters reports a continuing formal dialogue with another meeting planned. China’s own statement confirms that AI was discussed, providing first-party evidence that the subject moved beyond a unilateral US agenda item.
The story remains Confirmed for the occurrence of the talks and the US proposal. A mutually adopted incident mechanism is less settled because Beijing has not publicly described accepting it and no joint text has been released. The next decisive evidence would be a US Treasury or joint readout defining the channel, a Chinese statement confirming its status, or the planned Shenzhen meeting taking place with published terms or outcomes.