The United States and China are developing a direct alert channel for serious incidents involving each country’s superintelligence (SI) models, as officials pursue broader talks on shared AI risks, Treasury Secretary Scott Bessent said.
Bessent said the dialogue is intended to address risks shared by Washington and Beijing and reduce misunderstandings between the governments. The channel would provide prompt notices about serious incidents involving those models.
Bessent said the governments also plan to exchange best practices for evaluating advanced models and testing them before release. He said human oversight would remain part of those discussions.
In remarks released Nov. 11 by the U.S. Department of the Treasury, Bessent gave no start date for the channel or details on how alerts would be handled. He said he plans to accompany President Donald Trump to Shenzhen next week, where Trump is due to meet Chinese President Xi Jinping.
