US and China Agree to New AI Dialogue After Trump-Xi Summit
The US and China have wrapped up their dialogue, agreeing to continue formal talks on AI risks and benefits and create a channel for communicating about serious AI incidents in November. The agreement builds on a proposal made before the Trump-Xi summit to coordinate responses to AI incidents that could raise national security concerns.
Reading
In a Nutshell:
- The US and China will resume formal AI talks in November and establish a communication channel for serious AI-related incidents.
- Earlier discussions focused on risks from uncontrollable AI agents, cyber attacks and threats involving non-state actors, with US officials also calling for greater accountability from AI developers.
- The summit also produced an agreement covering $30 billion of non-sensitive goods in each direction, a new trade council and a two-month extension of the existing trade truce.
Context
The AI dialogue had already been taking shape before Xi Jinping's Washington visit. On September 20, Treasury Secretary Scott Bessent said Washington had proposed a system for notifying Beijing about serious AI incidents that could raise national security concerns. US Trade Representative Jamieson Greer said the mechanism would not cover Washington's controls on advanced AI chips or semiconductor manufacturing equipment.
Why It Matters
AI is becoming a regular part of US-China diplomacy, alongside trade and other strategic issues, with the two governments now discussing how to communicate when an AI system poses a serious security risk. The arrangement could eventually require agreement on what qualifies as a major incident, when governments should notify each other, and how they should respond, although those details have yet to be publicly established. The limits are also significant. Advanced chip controls were excluded from the earlier discussions, even though access to computing power is central to the development of increasingly capable AI systems, leaving the main technology competition largely outside the safety mechanism.
What we think
The presence of Jensen Huang, Tim Cook and Elon Musk at the dinner exposes the tension at the centre of this AI dialogue. The companies they lead have strong commercial reasons to keep advancing AI capabilities, while Washington and Beijing are trying to coordinate to address the dangers that increasingly capable systems could pose, including rogue agents and AI-driven security incidents. Bringing companies into the diplomatic process may give governments access to the expertise they need, but it also raises the risk of a conflict of interest that compromises the safeguards the dialogue is trying to build.