The US wants a hotline for AI disasters, and it just asked China to sign up.
Treasury Secretary Scott Bessent raised the idea during talks in Manhattan on Sunday, telling Chinese officials the two countries should notify each other when an AI system causes an incident serious enough to count as a national security event. Bessent was joined by US Trade Representative Jamieson Greer; on the other side of the table sat Chinese Vice Premier He Lifeng. No detail yet on what actually triggers a notification, how fast it would need to happen, or who verifies a claim once it's made. The proposal came up alongside broader trade discussions, not as a standalone AI summit.
This matters because it treats AI mishaps as the kind of event that could spiral into a state-to-state crisis, not just a product recall. It echoes the Cold War playbook: nuclear powers built direct communication lines specifically so a technical failure wouldn't get misread as an attack. The US and China already have a thin precedent here, having agreed in 2023 that AI shouldn't have a say in nuclear launch decisions.
An alert system is only as good as the definitions behind it, and right now nobody has said who decides what counts as a national security AI incident.