US, China experts seek AI nuclear ban before September 24 talks

The proposals from Melanie Sisson and Tianjiao Jiang arrive ahead of a planned meeting between Trump and Xi on September 24, giving the two leaders a concrete agenda item on AI and nuclear risk. They build on an agreement Biden and Xi reached in November 2024, but as of now binding mechanisms still don't exist, so the recommendations are advisory rather than enforceable.

Earlier efforts to address this issue include the National Security Commission on Artificial Intelligence (NSCAI), which called for preserving human control over nuclear weapons back in 2021. That precedent shows the concern is not new, yet the lack of binding rules suggests translating principles into practice remains difficult.

The proposal for a hotline for AI incidents highlights a specific risk: if a defensive AI system automatically responds to suspicious activity, the other side might read it as an attack. Carla Freeman of Johns Hopkins University questions whether such hotlines would work, pointing to China not picking up when the US called during the 2023 spy balloon crisis. Whether the September 24 meeting produces movement on any of these ideas is likely to depend on how both sides weigh the value of direct communication against recent experience.