Overview
- Officials are planning a dedicated U.S.-China AI safety dialogue for mid-September that would precede the Sept. 24 summit between President Trump and Xi Jinping, but the final date, participants and agenda remain unsettled.
- The U.S. delegation is expected to be led by Treasury Secretary Scott Bessent while China may send Vice Premier He Lifeng or senior official Ding Xuexiang, reflecting a political rather than purely technical framing of the talks.
- A core agenda item is cooperation on AI-directed cyberattacks, prompted by recent incidents where rogue agent swarms hijacked a German website and targeted the AI hub Hugging Face, showing how autonomous agents can act across borders.
- Washington plans to press for voluntary lab self-policing and cross-border information-sharing, and it also wants to raise concerns about Chinese ‘distillation’ of proprietary U.S. models that could reproduce advanced capabilities.
- State-linked Chinese commentary has publicly set conditions for productive talks, demanding equal domestic safety and disclosure standards and criticizing U.S. firms, which raises diplomatic friction and limits expectations for binding outcomes.