US Treasury Secretary Confirms Proposal for US-China AI Security Incident Notification Mechanism

US Treasury Secretary Scott Bessent confirmed to reporters on September 21 that the US has proposed a "notification mechanism" with China on AI: when AI-related incidents rise to the level of national security, the two sides would inform each other. According to WIRED, US and Chinese officials have begun negotiations on establishing a US-China AI Dialogue mechanism, which in addition to incident notifications would include regular meetings to align on shared goals and threats. Bessent said the dialogue could help form "a shared vision of common goals and common threats." This marks the first time the two countries have set up a dedicated bilateral negotiation channel specifically for artificial intelligence.

Confirmed

• Treasury Secretary Scott Bessent publicly confirmed the US has proposed an AI incident notification mechanism involving national security to China (m1, m4)
• According to WIRED, US and Chinese officials have begun negotiations on a US-China AI Dialogue mechanism covering incident notifications and regular meetings (m3)
• These are among the first bilateral dialogues between the two countries dedicated specifically to AI (m5)

Not Yet Confirmed

• Whether China will accept the proposal remains unclear, and the mechanism is still in an early proposal and negotiation stage (m1)
• A Reddit discussion thread (m2) notes that key questions remain unresolved: how to define an "incident"—for example, whether a model escaping its test environment, a major AI-assisted cyberattack, or dangerous capabilities discovered during training would count

Why It Matters

• This marks the first time AI safety governance has entered the level of US-China great-power coordination, establishing a communication channel analogous to the Cold War "nuclear hotline" (the analogy of the m2 author)
• If implemented, it would provide a direct communication mechanism for AI incidents and miscalculation risks, reducing escalation risks; Reddit author @yi111 believes having a communication channel is better than none, though the mechanism's definition and enforcement details remain the main challenges

2026-09-21 ~ 2026-09-21 · 5 related posts

• US proposes 'notification mechanism' for national security AI incidents in China AI dialogue — pstAsiatech · 2026-09-21
• US and China open first dedicated AI talks centered on a security incident notification mechanism — emmanuelvivier · 2026-09-21
• [source] US proposes AI incident notification mechanism with China, likened to a nuclear hotline — yi111 · 2026-09-21
• [source] US Treasury Secretary says US proposed an AI national-security notification mechanism to China — ShakeelHashim · 2026-09-21
• [source] US and China negotiate mechanism to alert each other on AI national security threats — nordicinst · 2026-09-21