In brief: The US and China aren't known for cooperating, but the United States wants to work with its rival to implement a safety notification mechanism for AI incidents. The news comes just days after reports that an AI hallucination almost started a war between the two countries.
Treasury Secretary Scott Bessent said that the proposal was part of weekend discussions ahead of talks at the White House taking place this week between President Donald Trump and China's Xi Jinping.
Bessent said the two countries had discussed establishing a new US-China AI dialogue, including a system for notifying each other of AI-related incidents with potential national security implications.
"We want a shared vision of common goals and common threats," Bessent said. "We think that just like any cross-border activity, that moving from opaque to more transparency between the No. 1 and the No. 2 AI powers in the world is very important."
Details of how the proposed notification system would operate remain unclear, including which agencies would manage it and what incidents would trigger an alert.
"The fact that both sides agreed to continue the dialogue is significant," George Chen, a partner and chair of digital practice at The Asia Group consultancy, said in written comments (via AP). "The initial outcome – an AI risk notification mechanism – sets a precedent that other countries may follow."
Separately, US and Chinese security experts have called for a dedicated military hotline to prevent accidental or unauthorized AI activity from being mistaken for a deliberate attack, though neither government has publicly endorsed that proposal.
Trump has repeatedly pushed back against any slowdown in AI development, calling safety concerns a "hoax." The president believes the US needs to stay ahead of China in the AI race for both economic and security reasons.
Incidents involving AI agents going rogue have dominated the headlines recently. But a potentially even more dangerous event took place earlier this year involving the military.
Sources say that the US came close to intercepting a Chinese vessel in the Middle East this spring after an AI-assisted intelligence report falsely suggested it was transporting materials for a nuclear weapons program. The claim originated in an inaccurate chatbot response.
According to two sources, armed military personnel were getting ready to board the ship and military aircraft were already airborne when officials reviewed the report more closely and found that AI had generated at least some of its content.