The United States and China plan to discuss testing advanced AI models and set up a direct channel for warning each other about serious incidents. The proposed Shenzhen dialogue puts risk management on the agenda even as both countries continue competing for AI leadership.
Watch Desk analysis
What happened
US Treasury Secretary Scott Bessent said the dialogue would advance at a meeting in Shenzhen alongside the Asia-Pacific Economic Cooperation summit on 18–19 November. He said the two sides expect to exchange best practices for evaluating and testing advanced models before release, and for keeping human judgement in control.
Bessent also said the countries had agreed to establish a direct channel for alerting each other to serious AI incidents. The Straits Times, citing his prepared remarks, reported that the arrangement is intended to reduce shared risks and prevent misunderstandings. The account describes plans and expectations, not a published agreement detailing how the channel will work.
Why it matters
A direct incident channel could give the two governments a way to communicate about serious AI problems while their broader relationship remains competitive. Sharing pre-release testing practices would also be a concrete form of cooperation, if the dialogue produces more than a meeting and carefully chosen language.
There are real limits to what is known so far. Bessent did not set out which incidents would qualify, how quickly either side would have to respond, or whether testing practices would be shared in detail. Those specifics will determine whether this becomes useful risk management or diplomacy with a promising name badge.
Our read
This is a notable attempt to put practical AI safety cooperation alongside strategic rivalry. The important test is not whether officials agree that serious incidents are serious, but whether they establish clear procedures for reporting them and act on what they learn.
What to watch
- Whether the Shenzhen dialogue takes place as planned alongside the APEC summit.
- What the incident-alert channel covers and how it will operate.
- Whether the countries publish details of any shared model-testing practices.
Discussion spark: Should the US and China share information about serious AI incidents even if doing so could expose weaknesses to a strategic competitor?
Sources and evidence
- Source update (11 November 2026, 01:28 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.