On October 6, 2026, OpenAI posted 722 AI-generated math preprints to a public GitHub repository, claiming progress on 372 open problems. A day later it withdrew three of them after finding an error, Retraction Watch reported. Many mathematicians say the release was rushed, and most of the results have not been independently checked.
This article reflects reporting available as of October 26, 2026.
What Happened
- October 6: OpenAI published 722 preprints in a GitHub repository, describing purported progress on 372 problems in geometry, computer science, algebra, and other areas. The papers came from an internal AI model.
- October 7: OpenAI announced it was withdrawing three manuscripts because of a sign error. The error in one paper carried into two others that built on it.
- OpenAI’s explanation: a spokesperson told Retraction Watch the company found the errors during an audit and described the process as iterative, as with other research manuscripts.
- Further changes: other coverage reports that OpenAI also revised 14 more manuscripts.
Why Most Results Weren’t Verified
A preprint is a research paper shared publicly before peer review. Formal verification goes a step further, with software checking every step of a proof.
The OpenAI spokesperson told Retraction Watch that the company’s collaborators at the Advisory Group on Mathematics and Artificial Intelligence recommended releasing the results without waiting for full formal verification, and that roughly half of the results were released unconfirmed.
How Mathematicians Responded
- Alex Townsend of Cornell said it was not surprising that one error spread to other papers, and he expects more errors to turn up.
- Andrew Sutherland of MIT called the quick withdrawal “the responsible thing to do,” while noting that many mathematicians remain unhappy with how OpenAI has handled its releases.
- Earlier criticism. After the release, a group of leading mathematicians voiced concern that OpenAI had rushed its earlier announcement about the Navier-Stokes equations, and Retraction Watch reports that more than 8,000 researchers have endorsed those concerns.
- A group statement quoted by Retraction Watch argued that releasing hundreds of files at once demonstrated power rather than scholarship, and urged mathematicians to stop working with OpenAI.
What This Does and Doesn’t Mean
The papers describe purported progress. They have not been independently confirmed, and three have already been withdrawn. That makes it inaccurate to say AI has solved hundreds of math problems.
The episode also shows a practical problem. An AI system can produce research-style papers far faster than people can check them. Critics argue the burden of reviewing falls on volunteer mathematicians, and checking a proof can take much longer than generating one.
What We Don’t Know
- How many of the remaining preprints will hold up under review.
- Whether more withdrawals or corrections will follow.
- Whether journals and preprint servers will set rules for large batches of AI-generated papers.
What to Watch Next
- Independent reviews by mathematicians of the preprints OpenAI says are verified.
- Further corrections or withdrawals.
- Whether OpenAI changes how it releases future results.
Bottom Line
OpenAI’s release is an early test of how research communities handle AI-generated work at scale. Until independent checks catch up, the results are claims, not established mathematics. At Future AI Wire, we’ll follow how the review process develops.
Frequently Asked Questions
Did OpenAI’s AI solve major unsolved math problems?
OpenAI claims progress on 372 open problems, but the results are not independently confirmed, and three of the papers have already been withdrawn.
What is a preprint?
A preprint is a research paper shared publicly before peer review. It can be useful, but it has not been vetted by independent experts.
Why did OpenAI withdraw three papers?
OpenAI said a sign error invalidated one argument, and two other manuscripts depended on it. The company said it found the problem during an audit.
Why are mathematicians upset?
Critics say releasing hundreds of papers at once, many without full verification, pushes the work of checking onto researchers who did not ask for it.
The Future AI Wire Editorial Team covers artificial intelligence, AI agents, agentic AI, and emerging technologies. We publish clear, research-informed explanations and practical insights to help readers understand AI developments, real-world applications, and the technologies shaping the future.