US and China Establish 'Super Intelligence Dialogue' and AI Incident Channel
The agreement marks the first formal bilateral AI dialogue structure between the two largest AI powers, with a dedicated channel intended to reduce the risk of miscommunication during serious AI incidents. The mechanism's usefulness depends on technical specifications that have not yet been agreed upon.
The United States and China agreed at a September 23-25 Washington summit between Presidents Trump and Xi Jinping to establish a formal bilateral AI dialogue and a dedicated communication channel for AI incidents. The US calls the arrangement the 'U.S.-China Super Intelligence (SI) Dialogue'; China calls it the 'China-U.S. AI Dialogue.' A follow-up exchange is scheduled by November 2026, and both sides agreed to sign a memorandum of understanding on military crisis communications. Chip export controls were explicitly left unchanged.
The agreement's practical limits are significant. No published bilateral technical specification defines what constitutes an incident, how notifications will be authenticated, or what response timelines apply. Insider sources told Politico the mechanism would be little more than an informal hotline. A functioning channel would require governments to specify notification thresholds, authenticated contacts, information-sharing procedures, and escalation rules.
The US proposed the mechanism during September 20 talks in New York between Treasury Secretary Bessent and Chinese Vice Premier He Lifeng. Bessent said the US sought 'a shared vision of common goals and common threats.' China's official readout acknowledged 'dialogues on AI-related issues' but did not confirm acceptance of Washington's specific notification mechanism before the summit.
The September 23-25 Trump-Xi summit produced a signed agreement naming the bilateral AI incident channel the "U.S.-China Super Intelligence (SI) Dialogue" and scheduling a formal follow-up dialogue for November.
A report published October 2 described a near-miss in which a US military analyst using AI misidentified cargo on a Chinese ship, bringing military planes into the air before the error was caught; sources called it an incident that "almost started a war."
Bloomberg Intelligence found Chinese AI models trail US counterparts by only 3% on benchmarks, down from 15% earlier in 2026.
On October 3, Washington rejected Anthropic CEO Dario Amodei's request for an antitrust waiver allowing AI labs to coordinate on development pace, with an official saying labs could already share safety information informally and that several had begun doing so.
Eric Schmidt argued in The Economist that any mutual AI pause is unverifiable: inspectors cannot read AI-generated code at scale and any party could cheat undetected. Legal analyses from Vinson and Elkins and Columbia Law outlined lawful coordination alternatives and warned a waiver could entrench dominant labs as de facto regulators.
Report described a US military analyst using AI who produced a false intelligence report misidentifying cargo on a Chinese ship, bringing military planes into the air before the error was caught; sources described it as having 'almost started a war'
A CSIS analysis of the September 24 Trump-Xi summit found that AI and chip policy received 'surprisingly muted' attention despite advance billing as a top-priority agenda item.
USTR Jamieson Greer confirmed no chip export control discussions took place, and China announced a bilateral AI dialogue as a post-summit outcome while the US issued no equivalent statement.
Confirmed outcomes from the September 24 Trump-Xi summit include an AI-specific dialogue scheduled for November 2026, an agreed MOU on military crisis communications, and future meetings at APEC in Shenzhen and the G20 at Trump National Doral Miami; one post-summit account described the incident channel as a "virtual red phone".
China is separately reported to be considering allowing Alibaba and ByteDance to import large quantities of Nvidia RTX Pro 5500 chips for AI servers, while the trade truce was extended only two months, leaving broader chip export controls unresolved.
Beijing reportedly considering allowing select Chinese companies to buy advanced Nvidia chips
Trump and Xi offered divergent AI governance stances at their September 24 summit: Xi said both countries have 'the capability and responsibility to develop and manage AI for good' and to keep it under human control, while Trump rejected international oversight as a 'globalist scheme' and named the DOJ as the US guardrail.
Politico's insider sources described the confirmed AI incident communication channel as little more than an informal hotline, with no published specification defining notifiable incidents, authentication, or response timelines.
Anthropic CEO Dario Amodei published 'We Must Pace the Frontier' on September 12, requesting an antitrust waiver so competing frontier AI labs can jointly discuss limits on development pace, drawing same-day endorsements from the heads of OpenAI, xAI, and Google DeepMind.
The September 23-25 Trump-Xi summit that produced the AI dialogue was also identified as including a $30 billion reciprocal tariff reduction, and the US formally named the agreed channel the 'U.S.-China Super Intelligence (SI) Dialogue'.
The White House confirmed a dedicated AI incident channel after Trump and Xi's September 24 meeting, while CSIS reported the summit's AI discussions were limited to shared safety standards, with USTR Greer stating chip export controls were not discussed.
China announced a bilateral AI dialogue with no timeframe; the US issued no parallel statement. CSIS identified a tension neither side addressed: both seek superior AI models while fearing those models could pose unmanageable threats, with three US developers reporting distillation attacks.
China's Foreign Ministry announced a November timeline for the first formal bilateral AI dialogue on September 29, and both sides agreed to a memorandum of understanding on military crisis communications, giving the incident channel from the September 24 Trump-Xi summit a near-term schedule.
China is also reportedly weighing letting Alibaba and ByteDance import Nvidia RTX Pro 5500 chips for AI servers, while the trade truce extended only through January 10 leaves chip export controls unresolved.
The Trump-Xi summit concluded September 25, with insider sources telling Politico the proposed bilateral AI notification mechanism, agreed at September 20 New York talks, would be little more than an informal hotline.
Xi said both countries have "the capability and responsibility to develop and manage AI for good" and keep it under human control, while Trump named the DOJ as the sole US guardrail. Senior AI executives including Altman, Zuckerberg, Pichai, Huang, Musk, and Bezos attended the state dinner.
At the UN General Assembly on September 22, Trump asserted the US leads China in AI 'by a lot' and rejected international AI oversight as a 'globalist scheme'.
The Trump-Xi summit followed as scheduled, with CSIS assessing minimal progress on the bilateral AI notification mechanism agreed at September 20 New York talks. China also publicly objected to US accusations of conducting hostile distillation attacks on its AI labs.
Ars Technica reported that the Justice Department accused six leading Chinese AI companies of IP theft two weeks before September 20 New York talks where Bessent proposed a bilateral AI incident notification channel, and experts warned the accusations alongside chip export controls may undermine cooperation.
China may view the safety dialogue as a cover for limiting its technological capabilities rather than a genuine shared-risk effort, the piece added. Beijing has not formally accepted the channel, which lacks any published technical specification.
The Trump-Xi summit produced minimal movement on AI governance, cyber operations, and export controls, per CSIS, following Trump's September 22 UN address in which he pledged to encourage rather than restrict superintelligence and designated the Department of Justice as the oversight mechanism.
AI labs seeking antitrust exemptions to coordinate on safety now face opposition from officials including Hawley, Vance, and Kratsios, who frame the request as a competitive reward rather than a coordination problem.
Sam Altman, Jensen Huang, Tim Cook, Mark Zuckerberg, and Dario Amodei were confirmed as expected state dinner attendees for Xi Jinping's Washington visit, and Treasury Secretary Bessent said the US is open to discussing AI 'shared risks' with China.
Huang publicly rejected Amodei's proposal for a broad ban on American technology sales to China. RAND published a paper describing the current US posture as analogous to accelerating with no investment in safety, and recommended preserving strategic options instead.
Perplexity CEO Srinivas argues export controls may make China a stronger long-term AI competitor by forcing mastery of physical infrastructure
PIIE published analysis arguing Trump-Xi summit unlikely to produce significant AI governance advances
Columbia Law analysis warned Amodei's waiver proposal could entrench dominant firms in AI governance architecture
Atlantic piece argues US rhetoric about existential AI race with China is undercut by lax export control enforcement
Foreign policy analysis argues US cannot credibly pressure China without domestic AI governance; describes neither side wanting to slow down first
Sanders calls on Trump and Xi to negotiate a treaty pausing AI development and banning superintelligence
US official Fick expressed interest in continuing AI dialogue with China, noting it could expand to include defense departments and arms control frameworks
Schmidt opposed a six-month AI pause, arguing it would benefit China.
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free