OpenAI Publishes 722 AI-Generated Math Proofs, Deepening Academic Controversy

Key points

• OpenAI published 722 manuscripts addressing 372 open mathematical problems, including the four-dimensional Kakeya conjecture and progress toward the Riemann hypothesis.
• The results were generated by an unreleased ChatGPT pioneer model, with an average of three hours of ChatGPT Pro use per result.
• The release includes reasoning summaries and compute estimates but lacks specific prompts and detailed compute times, contrary to AGMAI recommendations.
• The Institute for Advanced Study in Princeton criticized the practice, stating that AI can output arguments without human understanding or verification.
• OpenAI agreed to work with the Institute for Advanced Study but did not commit to stopping the use of advanced models for mathematical research.

Background

This release follows OpenAI's controversial solution to the Navier-Stokes Millennium Prize problem in September 2026, which sparked debates over transparency and attribution. The company also withdrew from a Caltech math hackathon amid backlash from mathematicians concerned about AI disrupting traditional research norms.

How outlets are covering it

Engadget highlights the scale of the release and the specific omissions in the data provided, noting that OpenAI ignored some AGMAI suggestions. The Guardian emphasizes the ethical concerns raised by the Institute for Advanced Study, focusing on the lack of human verification and the risk of a two-tier system where AI labs outpace the broader mathematical community. Both sources agree that the results require rigorous assessment by mathematicians before their impact can be understood.

Why it matters

The release underscores the growing influence of AI in high-level mathematics and the tensions between rapid AI-driven research and traditional academic rigor. It raises questions about transparency, verification, and the equitable access to advanced AI models, potentially reshaping the future of mathematical research and collaboration.

What to watch

Mathematicians will assess the validity of the 722 manuscripts, and OpenAI may face further scrutiny from academic institutions. The company's collaboration with the Institute for Advanced Study could lead to new guidelines for AI-driven research, but the debate over transparency and access is likely to continue.