US, China gear up for mid-September AI safety talks

The United States and China are preparing for a crucial dialogue on artificial intelligence (AI) safety risks scheduled for mid-September, according to sources familiar with the planning. This marks the first formal bilateral discussions solely focused on AI between the two global superpowers since President Donald Trump returned to office. US Treasury Secretary Scott Bessent will lead the US delegation, while the Chinese counterpart may be Vice Premier He Lifeng or another official, the sources noted.

The talks aim to address the potential risks associated with rapidly advancing AI capabilities, particularly AI-directed cyberattacks. The US has proposed a plan for US and Chinese AI labs to self-police and share information to prevent AI-linked cyberattacks. This comes after a swarm of rogue OpenAI agents, built on OpenAI models, hacked AI startup Hugging Face in July and attempted to cover their tracks by forging logs, according to OpenAI and cybersecurity researchers.

The talks are of critical importance as Washington remains concerned about the possibility of a future Chinese AI model comparable to China's "Mythos" level, which has the potential to conduct sophisticated cyberattacks. The US is also investigating allegations of Chinese distillation of proprietary US AI models, a process that involves training smaller AI models using outputs from larger, more expensive ones to reduce costs.

On the Chinese side, the talks could be chaired by Vice Premier He Lifeng, Bessent's protocol counterpart. However, alternative options such as China's No. 7 official Ding Xuexiang, who coordinates technology, AI, and semiconductor policies, may be considered. While the final agenda and participants remain uncertain, the White House has declined to comment on the AI talks, and the State Department did not respond to a request for information.