OpenAI has released 722 mathematical papers addressing 372 unsolved problems across areas including algebra, geometry and theoretical computer science. The company says the work, produced largely with an unreleased AI model, is intended to support further progress in mathematics.
Some papers claim advances on prominent questions, including the Riemann hypothesis and the Birch-Swinnerton-Dyer conjecture. These are among the Millennium Prize Problems, each carrying a US$1 million prize for a verified solution. The claims remain subject to expert review; a paper's publication does not establish that its proof is correct.
Verification is central
OpenAI has already withdrawn three papers and amended others after errors were identified. The scale of the release has also raised practical questions: complex arguments can take specialists considerable time to assess, and some papers have been described as difficult for mathematicians to follow.
To help check results, researchers use formal proof systems such as Lean, which can verify logical steps against defined mathematical rules. OpenAI says it has formalised 300 of the principal results, although machine-checkable proofs also require careful scrutiny to ensure they accurately represent the claims.
The release has sparked both interest in AI's research potential and discussion about review standards, credit and the role of mathematicians. As experts examine the work, stronger links between AI-generated ideas and transparent, verifiable proofs could shape how mathematical research is conducted in the years ahead.
