US and China Agree to an AI Incident Hotline, Not a Slowdown
After an eight-hour negotiating session at JPMorgan Chase's New York headquarters on September 20 and a three-day Trump-Xi summit in Washington that concluded September 26, the United States and China agreed to establish a bilateral communication channel for AI-related incidents, with a dedicated AI dialogue set for November, according to PBS NewsHour and Al Jazeera reporting corroborated by statements from Treasury Secretary Scott Bessent and White House science adviser Michael Kratsios. The September 20 session paired Bessent and US Trade Representative Jamieson Greer with Chinese Vice Premier He Lifeng and chief trade negotiator Li Chenggang.
The mechanism is explicitly modeled on Cold War-era crisis channels rather than any new regulatory body: Al Jazeera traces the lineage to the 1997 secure phone line established under Presidents Clinton and Jiang Zemin and the 2008 military hotline, which carries a 48-hour response requirement that US officials say Beijing has sometimes left unanswered. Bessent described the goal as moving the two powers "from opaque to more transparency," and Greer specified that export controls on advanced AI chips and semiconductor manufacturing equipment were deliberately excluded from the arrangement. No party has publicly defined what threshold of incident would trigger a notification, and China's Xinhua news agency has so far only acknowledged that discussions took place, stopping short of confirming the mechanism itself.
The agreement lands squarely against a week of public catastrophic-risk warnings from the industry's own leadership. Anthropic CEO Dario Amodei published an essay urging AI developers to "pace the frontier" and calling for the US and China to negotiate an agreement to substantially limit the rate of frontier AI development; OpenAI CEO Sam Altman separately pressed for a slowdown and stronger regulatory oversight. Neither proposal resembles what the two governments actually signed off on.
"We need an immediate, indefinite international moratorium on frontier AI development." — David Krueger, AI safety specialist, University of Montreal
Both governments rejected the premise behind those calls rather than negotiating over it. Al Jazeera reports the Trump administration has dismissed AI existential-risk advocacy as a "conspiracy" and vowed to oppose any "globalist scheme" to constrain US AI development, with Trump telling reporters the US is "leading China by a lot" and intends to keep it that way. China, for its part, released the third iteration of its own AI security framework in the same window but characterized existential-risk concerns as "fearmongering," framing its own document around geopolitical and domestic-control risks instead. Existential Risk Observatory director Otto Barten's warning that "we do not know where the positive feedback loop of ever-smarter AIs creating ever-smarter AIs without human supervision will end" went unanswered by either side's public statements.
The practical signal for the industry is that state-level AI diplomacy is consolidating around crisis containment, not prevention. Carnegie Endowment's Scott Singer called the notification channel "an obvious win for both sides" precisely because it asks so little of either government — it creates a reporting line for incidents after they occur rather than any constraint on the compute, capability thresholds, or deployment pace that safety researchers argue drive the underlying risk. With chip export controls carved out and trigger conditions undefined, the channel's design tracks closer to the 2008 hotline's uneven record than to a binding safety commitment, leaving investors and lab safety teams with the same unconstrained scaling trajectory they had before the summit — now with a phone line attached.