Experts from the US and China are advocating for international regulations to prohibit autonomous AI systems from managing nuclear weapon deployments. Melanie Sisson from the Brookings Institution and Tianjiao Jiang of Fudan University released these recommendations prior to a scheduled meeting between Donald Trump and Xi Jinping on September 24.
The proposed guidelines establish specific restrictions, including a ban on AI systems launching nuclear weapons or targeting nuclear command infrastructure. The experts emphasize that human oversight must remain mandatory for any AI-assisted cyberattacks on strategic systems. Furthermore, they suggest that both nations establish a unified definition of human control.
To mitigate risks, Tianjiao Jiang suggests creating a dedicated hotline for AI-related incidents. This communication channel aims to prevent accidental escalations if defensive AI systems react to suspicious activity. The proposal expands on a previous agreement reached by Joe Biden and Xi Jinping in November 2024.
Despite these efforts, binding international mechanisms remain absent. Previous calls for human control, such as the 2021 recommendation by the National Security Commission on Artificial Intelligence, have not resulted in formal agreements. Skepticism persists regarding the efficacy of hotlines, citing the 2023 spy balloon incident when communication channels failed.
