- Scott Bessent said the United States proposed notification of AI incidents affecting national security.
- The proposal arose in talks with Chinese Vice Premier He Lifeng.
- A proposal does not establish an operational bilateral alert system.
An incident channel enters the talks
US Treasury Secretary Scott Bessent said Washington had proposed a mechanism for notifying China about AI incidents with potential national-security implications, the Associated Press reported on September 21. He described the proposal after talks with Chinese Vice Premier He Lifeng in New York, ahead of planned presidential discussions.
The announcement concerns communication about dangerous incidents. It does not amount to a joint limit on model development, a shared approval process for AI products or an agreement to exchange model weights.
The design questions are consequential
An incident channel would need a usable definition of an incident. A system failure inside one company, malicious use of an otherwise functioning model and an event affecting infrastructure across borders would create different demands for notification. The announcement reviewed here does not establish where those boundaries would sit.
Timing would matter too. An early warning might contain incomplete information, while a fully investigated report could arrive too late to help the other side respond. A practical mechanism would need designated recipients, a way to authenticate messages and a process for updating an initial assessment. These are implementation questions, not details that have been announced.
How to read the diplomatic signal
The proposal indicates that competition and discussion of shared AI risks can appear in the same diplomatic agenda. It should not yet be described as an operating hotline or a settled safety pact.
For readers following the next stage, the useful evidence would be a joint statement or agreed terms identifying participating agencies, covered events and the point at which notifications begin. Until then, the confirmed development is that the United States put an incident-notification proposal into the talks; Chinese acceptance and the final design remain unestablished in the reporting cited here.
Sources & context
Go to the original material. Company claims remain attributed to their sources.
01Updates & corrections
— Expanded with source reporting, context and a clearer distinction between announced plans and demonstrated results.
