OpenAI posts 722 AI maths preprints, three pulled next day

OpenAI uploaded 722 AI-generated mathematics manuscripts to GitHub on 6 October 2026. Three were withdrawn a day later after a sign error invalidated one proof and the two papers that depended on it, Nature reported; Retraction Watch logged the pulls the following day.

The papers came out of an unreleased internal model pointed at roughly 4,000 problems. OpenAI claims progress on 372 open questions across geometry, algebra and computer science. None of the five remaining Millennium Prize Problems are in the batch.

Alvaro Lozano-Robledo, a number theorist at the University of Connecticut in Storrs, called the simultaneous drop "staggering" and asked, "Why is OpenAI trying to solve hundreds of math problems to be released all at once?"

Ursula Martin, a mathematician at the University of Oxford, described the strategy as "tossing the community a messy first draft and expecting us to clean it up for them." Michael Harris, a number theorist at Columbia University in New York City, told Nature he "can't bring myself to spend my time studying" the preprints and called the publication process "illegitimate."

Not every voice was hostile. Levent Alpöge, a mathematician at rival lab Anthropic in San Francisco, posted on X that whatever one thinks of the release, "it's obviously the most significant moment in mathematical history."

An OpenAI spokesperson said the company aims to "provide tools for mathematicians and scientists" and evaluates its models "against difficult frontier problems." Separately, mathematicians raised the worry of being scooped on research programmes they had been running for years. Three of the researchers we track at AI Weekly posted the Nature piece on day one.