What happens to nuclear deterrence when superintelligent AI automates military research? According to a new essay on LessWrong, the answer is concerning. Historically, nuclear deterrence has weathered new military technologies—ICBMs, multiple warheads, thermonuclear weapons—because progress was gradual enough for rival states to adapt. But AI-driven development could compress that timeline. A power with a technological lead through superintelligence might overcome deterrence three ways: by locating and destroying all enemy nuclear forces before retaliation becomes possible; by building defenses that neutralize retaliatory strikes and alternative delivery methods; or by using superior intelligence to manipulate adversaries' decision-making. The analysis concludes that deterrence will likely hold—but the margin of safety is narrowing. Without comparable AI, even nuclear-armed states could eventually be outpaced.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton