US Treasury Secretary Scott Bessent said on Monday, 21 September 2026, that senior officials from the United States and China will hold a new meeting in the Chinese city of Shenzhen in about two months to discuss artificial intelligence risks and communications protocols when incidents related to technology safety occur.
Washington and Beijing establish channel for artificial intelligence incidents
Bessent said Washington and Beijing had agreed to formalise a bilateral dialogue on artificial intelligence, including the establishment of a dedicated communications channel for sharing information about incidents that could raise safety concerns or reach a level affecting national security.
The agreement came during talks held in New York on Sunday, 20 September, between Bessent and Chinese Vice-Premier He Lifeng, who led his country's delegation, with US Trade Representative Jamieson Greer also taking part. The two sides discussed economic and trade issues as well as matters related to artificial intelligence.
Risks from uncontrolled agents and cyber threats
Bessent said the next phase would involve beginning discussions on protocols that would allow the two sides to reach a common understanding of the main risks posed by artificial intelligence. He said these risks could include agents capable of operating without control, as well as threats from non-state entities or non-governmental actors active in cyberspace.
He added that establishing an incident-response channel was intended to enable communication between the two countries when an artificial intelligence-related incident arises that warrants a warning or the exchange of information, as Washington and Beijing seek to identify common risks and mechanisms for dealing with them.
China confirms constructive dialogue on technology and trade
The Chinese government, for its part, said the two countries' teams held candid, in-depth and constructive talks in New York on economic and trade issues of common interest and on implementing understandings reached in previous rounds. It confirmed that the delegations also held a dialogue on issues related to artificial intelligence.
Bessent also addressed the responsibility of developers of artificial intelligence systems for their products, reiterating his view that laboratories and companies must bear the consequences of what they develop, including cases in which systems or agents operate without control and generate uncontrolled hacking activity.
Speaking about an incident involving Hugging Face's open-source tools platform, Bessent held OpenAI's management responsible, saying that responsibility lay with human management rather than a group of agents. He added that artificial intelligence laboratories were required to take responsibility and could slow the pace of development whenever they wished.
Bessent described trade relations between the two countries positively, saying that the stability of the relationship since last autumn had been a central focus of the talks and expressing his belief that both sides would maintain that stability. The expected Shenzhen round will focus on artificial intelligence risks and rules for communication and incident response.
