Arms Control Association executive director Daryl G. Kimball examines how major powers’ integration of AI into nuclear command-and-control systems is creating new risks of miscalculation and unintended escalation beyond traditional deterrence uncertainties. He argues that the ‘human in the loop’ principle currently lacks agreed definitions, enforcement mechanisms, and sustained international dialogue among nuclear-armed states. Kimball calls for binding international agreements guaranteeing human control over nuclear-weapons decisions and for technical safeguards to prevent AI-enabled escalation.
