OpenAI says its AI has refuted a 30-year-old computer science hypothesis as experts rush to check the claim

OpenAI disclosed on 26 October that its internal AI models had generated three findings touching on long-standing open questions across computer science, mathematics and physics. The computer science finding — a refutation of the Exponential Time Hypothesis — is drawing the most pressing outside scrutiny, with independent researchers cautioning that verification would fundamentally alter theoretical computer science. Mathematicians and physicists are separately examining the other two claimed proofs, neither of which OpenAI has made fully available for public review.

The Exponential Time Hypothesis is a decades-old conjecture on which hundreds of studies into computational limits rest — and overturning it could undermine the foundations of modern encryption.