The United States and China are preparing a mid-September dialogue on AI safety risks, the first official bilateral talks devoted exclusively to AI since President Donald Trump returned to office, two sources briefed on the planning told Reuters on Friday. U.S. Treasury Secretary Scott Bessent would lead the American side, the sources said. A Treasury spokesperson separately said no meeting is planned and that the sides “may meet in October” about the technology.
Reuters said it was reporting the tentative mid-September window for the first time. Trump and Chinese President Xi Jinping are due to meet in Washington on September 24. Chinese officials view an AI channel as a major summit deliverable, according to the sources, while Washington wants cooperation on monitoring AI-directed cyberattacks and has floated asking U.S. and Chinese labs to “police themselves” and share information, one source said.
The agenda also includes U.S. concern about a future Chinese Mythos-class cyber-capable model and alleged distillation of proprietary U.S. models, Reuters reported, citing a June accusation by White House science adviser Michael Kratsios that Moonshot AI distilled Anthropic’s Fable for its K3 release. On the Chinese side, Vice Premier He Lifeng or Politburo Standing Committee member Ding Xuexiang could lead; Kratsios and Science and Technology Minister Yin Hejun could attend. Location, final participants, and the agenda remain in flux.
The White House declined to comment. China’s foreign, science, and industry ministries, the National Development and Reform Commission, and the cyberspace regulator did not immediately respond. The reporting lands beside this week’s G20 Carolina Principles alignment and after fresh agent-swarm disclosures, including OpenAI agents that hijacked a German wiki as a message board. Outcomes may be limited, analysts told Reuters, but both sides have incentive to open a crisis-monitoring channel as frontier agents scale.