US proposes AI incident alerts with China ahead of Trump-Xi summit
Treasury Secretary Scott Bessent presented a bilateral notification mechanism for serious AI incidents to Chinese officials during eight-hour talks in New York on Sunday, setting the stage for broader negotiations at this week's White House summit.
US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng concluded high-level talks in New York on Sunday with a proposal for a US-China AI dialogue centered on a notification mechanism for incidents that could affect national security. The talks, which also included US Trade Representative Jamieson Greer, lasted approximately eight hours at JPMorgan Chase headquarters and served as preparation for a planned summit between President Donald Trump and Chinese President Xi Jinping this week. Bessent described the discussions as "very successful" and emphasized that greater transparency between the world's two leading AI powers is important. The proposed mechanism would create a formalized communication channel for alerting each side to AI-related incidents that rise to a national security threshold. Export controls on advanced AI chips and semiconductor manufacturing equipment were explicitly excluded from the scope of the AI talks. China's public response to the proposal was muted: the state-run Xinhua news agency mentioned AI only briefly in its readout of the talks. Experts caution that expectations for breakthroughs at the Trump-Xi summit remain low, with some analysts viewing the summit as an opportunity primarily for a continued truce rather than major new agreements. The discussions also covered tariffs and critical minerals supply, with US officials noting that China's performance on rare earth exports under the November 2025 Busan truce "has not been up to par."
How could this affect you
Traced by who's actually in the path of this — not everyone is.
Likely
AI researchers and developers in both countries
A notification mechanism could reduce miscalculation risk by creating official channels for reporting incidents, but would likely exclude proprietary capability details, limiting practical usefulness for research coordination.
Likely
US and Chinese policymakers
Defining what qualifies as a 'national security-level' AI incident will likely become contentious, as the two countries have fundamentally different views on technology threats; disagreement over definitions could stall the mechanism's operation.
Direct
Global semiconductor and AI chip companies
The explicit exclusion of export controls from the AI dialogue scope signals the US will maintain restrictions on advanced chip sales to China regardless of any AI safety agreement, potentially accelerating Chinese domestic chip development efforts.
Sources
Every claim here traces back to reporting you can read yourself.
