The Baruch Plan: A 1946 Lesson for Today's AI Governance Debate

In 1946, following the atomic bombings of Hiroshima and Nagasaki, the US (then the sole nuclear power) proposed the Baruch Plan at the United Nations. The plan envisioned an international Atomic Development Authority that would own all uranium and thorium resources, from mining to production, and institute strict inspections. The US would eventually dismantle its own arsenal. Despite initial Soviet openness, the USSR rejected the plan—fearing loss of veto power, intrusive inspections, and the chance to build its own bomb. By 1949, the USSR had detonated its first atomic weapon, and the window for international control closed. The UN Atomic Energy Commission dissolved in 1952, replaced by a weaker IAEA in 1957.

Today, the same dynamic replays around advanced AI. Leading scientists warn of existential risk; governments propose pauses or international oversight. Yet rival powers—fearful of falling behind—resist constraints. The Baruch Plan illustrates how technical superiority + early diplomatic initiative can be thwarted by geopolitical mistrust. For AI, the lesson is stark: without binding, verifiable agreements that address power asymmetries, uncoordinated races will likely dominate over cooperative safety regimes. The plan's failure is a cautionary tale for AI governance today.