Autonomous AI threats leave US and China minutes to prevent nuclear crisis, experts warn

US and Chinese security experts warn that autonomous AI systems could trigger a nuclear crisis, leaving both nations with minutes to respond to perceived attacks. These experts, from a US-China dialogue, propose several measures to prevent such scenarios, including:

• Establishing "red lines" around nuclear systems.
• Ensuring human control over significant cyberattacks.
• Creating a dedicated hotline for AI-related incidents.

Melanie Sisson of the Brookings Institution emphasized that humans should retain sole authority for AI-enabled cyberattacks on critical infrastructure. Tianjiao Jiang of Fudan University advocated for explicit red lines preventing AI from independently deciding to use nuclear weapons or attack nuclear command systems, along with a shared definition of "meaningful human control." Jiang also proposed a military hotline for AI incidents, noting that automated cyber exchanges could escalate rapidly. However, the utility of existing hotlines has been questioned due to past communication issues, such as China's refusal to answer US calls during the "spy balloon incident."