US–China: From "No AI Guardrails" to an AI Incident Channel and a November Dialogue
Update (Sept 26–27): the working-level follow-ons landed two days after the summit. Per AP, in statements issued Saturday, the US and China agreed to:
- set up a communication mechanism for AI-related incidents, "to discuss related risks and benefits,"
- hold an AI-specific dialogue scheduled for November,
- sign a memorandum of understanding on "strengthening crisis communications and preventing crises between the two militaries,"
- keep working through a Board of Trade (the White House says it went into operation this week), cut reciprocal tariffs on roughly $30B of goods, and extend the trade truce by two months.
This is a delta from the summit itself: the Sept 24 Washington summit produced a trade truce extension but no AI guardrails, with Trump saying "I want to leave it exactly where it is." The Saturday statements now add an institutional AI channel — modest, but the first bilateral US–China AI incident mechanism of this administration. Analyst Wang Zichen (Center for China and Globalization): "the emphasis on working-level engagement, military-to-military dialogue and crisis-management mechanisms suggests an effort to make that stability more durable and institutionalized."
What it means: the channel covers incident communication and risk discussion, not frontier-model safety standards — but it is the kind of institutional hook the labs' UN Security Council push (Labs Take the Pacing Case to the UN Security Council — and the US Rejects Global AI Governance On the Spot) failed to get, and it sits alongside the White House telling OpenAI and Anthropic to withhold new models from the UK's AISI (White House Tells OpenAI and Anthropic to Withhold New Models From UK's AISI Until the US Reviews Them First). Washington is building a direct Beijing channel on AI risk while bottlenecking allied evaluation access. Next milestone: the November AI dialogue, alongside APEC in Shenzhen.