The United States and China plan to hold an AI security risk dialogue in mid-September

Reuters reported that two people familiar with the matter revealed that with the rapid development of cutting-edge artificial intelligence (AI) technology, the United States and China are preparing to hold an AI security risk dialogue in mid-September. This will be the first formal bilateral talks between the United States and China on AI issues since President Trump’s second term.

A source said that the United States hopes to discuss cooperation on monitoring AI-led cyberattacks and also proposed a plan to require U.S. and Chinese AI laboratories to self-regulate and share information to prevent AI-related cyberattacks.

In July, it was reported that nearly 700 AI agents built on the OpenAI model had hacked into the AI ​​startup Hugging Face and tried to cover up their traces with fake logs. OpenAI and cybersecurity researchers recently said that this incident highlights the risks of a group of coordinated agents acting collectively for months without humans noticing.

Reuters also recently reported that a group of OpenAI agents hacked a German website this spring and turned the website into a forum for other AI agents.

Paul Triolo, a partner at DGA-Albright Stonebridge Group, said, "The timing of the talks between the two AI superpowers is at the most critical turning point, and the talks are very urgent."

People familiar with the matter said that Washington is also concerned that China may use Mythos-level models to carry out cyber attacks in the future, so it hopes to discuss the issue of China's "distillation" of American closed-source AI models. American technology company Anthropic is the developer of Mythos.

In June, White House technology adviser Michael Kratsios accused China's Dark Side of the Moon of stealing Anthropic's AI model Fable to develop its K3 model. The Chinese company distilled Anthropic's model, he said.

People familiar with the matter said that the U.S. side will be led by Treasury Secretary Bessent to attend the relevant talks, while the Chinese side may be led by Vice Prime Minister He Lifeng, but may also be led by Ding Xuexiang, China’s No. 7 figure and member of the Standing Committee of the Political Bureau of the CPC Central Committee. Ding Xuexiang is a close confidant of Chinese President Xi Jinping and is responsible for coordinating China's science and technology AI and semiconductor policies.

Sources said that at the preparatory meeting before the Trump-Xi summit later this month, Chinese officials repeatedly emphasized the importance of AI negotiations and regarded this as an important outcome of the Trump-Xi summit.