OpenAI drops 722 AI maths preprints, mathematicians push back

OpenAI posted more than 700 AI-generated mathematics preprints to GitHub in a single drop on 6 October, produced by an unreleased model, Nature reports. The batch did not solve any of the Millennium Prize Problems.

Reaction from named mathematicians ran from refusal to disbelief. Alvaro Lozano-Robledo, a number theorist at the University of Connecticut in Storrs, asked on X: "Why is OpenAI trying to solve hundreds of math problems to be released all at once?" Michael Harris, a number theorist at Columbia University, told Nature he considered the publication illegitimate: "I can't bring myself to spend my time studying [these preprints], so I will not be reading these papers."

Ursula Martin, a mathematician at the University of Oxford, said the release amounted to "tossing the community a messy first draft and expecting us to clean it up for them."

Not everyone recoiled. Levent Alpöge, a mathematician at rival lab Anthropic in San Francisco, wrote on X that "it's obviously the most significant moment in mathematical history," while also flagging the "sad stories" waiting for junior researchers who may be scooped on problems they have worked on for years.

The audit problem showed up fast. A day after the drop OpenAI withdrew three of the 722 manuscripts and revised 14 more after an internal review found a sign error that invalidated an argument in one paper and the two others that depended on it, Retraction Watch reported. Alex Townsend, an associate professor at Cornell, told the publication: "Given the skepticism of AI in the mathematics community, I think that OpenAI should have announced the manuscripts that were lean verified first." Andrew Sutherland, an MIT senior research scientist, called the swift withdrawal "the responsible thing to do," but added that mathematicians remain "unhappy with how OpenAI has behaved up to this point."

OpenAI told Nature it tests its models against "difficult frontier problems in all sciences." Two researchers we track posted the Nature piece in the days after the release. The repository remains live at github.com/openai/math.