OpenAI’s Mass Release of AI-Generated Math Solutions Sparks Global Debate
On October 6, 2026, OpenAI published over 700 AI-generated manuscripts to GitHub, claiming solutions to more than 370 outstanding mathematics problems, including progress on three of the seven Millennium Prize Problems. The papers were produced by an unreleased internal model, averaging roughly three hours of compute per solution. The New York Times
The move has split the mathematics community. Supporters like University of Toronto’s Dan Litt praised the release as “great for mathematics,” welcoming the open GitHub format. Fields Medal winner Terence Tao declared the “Math 1.0” era over, calling for a new community-driven “Math 2.0.”
Critics have been sharp. The Association for Human Mathematics issued an open letter calling the mass release “a demonstration of power, not scholarship,” urging mathematicians to cut ties with OpenAI. NYU’s Tristan Buckmaster raised plagiarism concerns, saying OpenAI had not done enough due diligence. Within days, OpenAI quietly withdrew solutions to at least three problems after errors were found.
Verification remains the central challenge: checking hundreds of complex proofs falls on unpaid mathematicians, and many papers lack complete machine-checked formalizations. OpenAI says it will fund workshops and release further compute documentation.
Read our previous article on AI related news here.
