US-China AI Hotline Talks Progress, But Deal Months Away
Original: A US-China AI Hotline Won't Be Ready For a While
Why This Matters
A bilateral AI incident-notification system would be the first formal US-China mechanism for managing frontier AI risks at a national security level.
Trump and Xi met for a state dinner in Washington on Sept 25, 2026, with a proposed US-China AI notification system—modeled on Cold War-era hotlines—topping the agenda. Treasury Secretary Bessent announced the concept Sunday, but officials say a formal agreement is still weeks away, targeting finalization before year's end.
The Trump-Xi state dinner brought AI front and center, with OpenAI and Nvidia executives confirmed attendees. Treasury Secretary Scott Bessent introduced the AI notification system concept on Sunday alongside Chinese Vice Premier He Lifeng—but enthusiasm at the top hasn't translated into a signed deal.
Trump administration officials told Wired's Inner Loop the framework won't be finalized for at least several weeks. The sticking points are significant: no agreed benchmarks on what constitutes a national security-level AI incident, and China hasn't formally committed to the arrangement. The administration is targeting a functional framework by Bessent's follow-up meeting with Lifeng in Shenzhen, roughly two months out.
Enforcement is another headache. Officials want reciprocal obligations—not just the US alerting China—but with Trump extending a trade truce and reluctant to halt Nvidia chip exports, any mechanism with real teeth faces internal resistance. The US has already granted limited licenses for Nvidia's H200 chip sales to China this year.
Adding another variable: Trump hasn't named an 'AI czar' to lead a planned interagency 'AI Force.' Bessent is one name in circulation, alongside Commerce Secretary Howard Lutnick and White House AI officials Michael Kratsios and Sean Cairncross. The vacancy itself may further slow talks.