Proposal Calls for a Permanent U.S.-China Strategic AI Dialogue

Summary

The article presents a proposal for a permanent, senior-level Strategic AI Dialogue (SAID) between the United States and China to reduce risks from increasingly capable artificial intelligence. It recommends Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng lead the mechanism, with the White House Office of Science and Technology Policy, led by Michael Kratsios, serving as a standing U.S. science and technology participant. SAID would meet quarterly in a neutral country, preferably Switzerland, with principals, senior officials, and technical sessions supported by continuous working groups. The proposal is explicitly designed as a risk-reduction channel, not a comprehensive AI-governance negotiation or a way to settle disputes over chips, sanctions, investment restrictions, Taiwan, or market access. Its initial agenda includes keeping humans in control of nuclear-use decisions, reducing catastrophic biological risks such as mirror-life research, developing comparable frontier-model evaluations, and creating a hotline for severe cross-border AI incidents. It also proposes cooperation on AI-related financial instability, employment research, high-consequence cyber operations, transnational fraud, terrorist access to dangerous capabilities, and civilian autonomous-system safety. The suggested first-year deliverables include tabletop exercises, common terminology for dangerous model capabilities, emergency contacts, an employment assessment, and possible bilateral statements on nuclear authority and mirror organisms. The author ranks employment, mirror life, nuclear human control, fraud, and civilian-system safety as the most feasible starting points, while treating cyber operations, conventional military AI, and loss-of-control scenarios as harder longer-term subjects. The central argument is that mutual vulnerability can support limited cooperation even without strategic trust or an end to U.S.-China technological competition.