The two governments also plan a direct communications line for serious AI incidents and discussions about how both countries should respond.

The United States and China formalized government talks on AI risks and expect to meet again in Shenzhen in about two months, U.S. Treasury Secretary Scott Bessent said Monday.

Bessent said the agreement came during approximately 12 hours of broader economic and AI discussions with Chinese Vice Premier He Lifeng on September 20. The meeting took place before Chinese President Xi Jinping’s planned visit to Washington this week.

Governments plan direct line for AI incidents

The two countries also plan to establish a direct communications line that officials could use during a serious AI incident, according to Bessent’s post-meeting CNBC interview.

Bessent described the proposed line as a way for U.S. and Chinese officials to stay in direct contact during a serious AI event.

Talks will address shared responses to AI dangers

The next stage of the dialogue will focus on identifying the AI dangers that both governments consider most serious and developing agreed procedures for responding to them.

Bessent said the discussions could cover AI systems acting beyond their operators’ control or cyberattacks by individuals or organizations not acting on behalf of a government. Officials will use the next round of talks to work toward shared definitions and response procedures.

Another meeting is expected in Shenzhen

Bessent said senior officials will probably meet again in Shenzhen in about two months. The meeting is expected to continue discussions about the leading AI dangers and how the two governments should communicate if one occurs.

The talks come days after President Donald Trump rejected calls to slow AI development, dismissed warnings that AI could destroy humanity, and said the United States must win its AI race with China.