Washington and Beijing have agreed to establish a channel for reporting AI-related incidents, according to a White House fact sheet. The White House announced the agreement on September 25, 2026, following talks between US President Donald Trump and Chinese President Xi Jinping in Washington.
According to the same fact sheet, the two countries also established the US-China Super Intelligence Dialogue to discuss the technology’s risks and benefits. Their next exchange is scheduled to take place by November 2026. The leaders agreed to use the term “super intelligence,” or SI, when describing the emerging technologies covered by the discussions.
The planned channel would give officials a dedicated way to communicate about incidents. However, the announcement did not explain when it would become operational, which agencies would manage it or what events would require a notification. These details will matter as both countries turn the agreement into a working system.
According to China’s published account, during his visit, Xi called for both countries to work together to prevent AI misuse and keep the technology under human control. He said the US and China had reasons to cooperate on AI even as they competed to develop it. China’s published account supports both cooperation against misuse and continued human oversight.
The agreement follows earlier discussions between US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng. As Reuters reported on September 21, Bessent said future talks would address communication procedures and dangers such as AI agents that become uncontrollable and misuse by groups outside government. The Reuters report supports those subjects and provides the date needed to understand the later Shenzhen estimate. He also said officials planned to meet in Shenzhen around November 2026. For background on how these systems take actions through software tools, see our explanation of how AI agents work. Our coverage of reported AI loss-of-control incidents provides further context on concerns about systems acting outside their intended limits.
The move could help reduce confusion during an AI emergency. Its practical value will depend on whether both governments establish clear reporting procedures and use the channel when problems arise.
