OpenAI released hundreds of solutions to open mathematics problems despite a promise to mathematicians, while leaked meeting records showed an employee had threatened a researcher’s career
OpenAI released hundreds of solutions to open mathematics problems despite a promise to mathematicians, while leaked meeting records showed an employee had threatened a researcher’s career
On October 6, 2026, OpenAI published 722 manuscripts presenting solutions to open mathematics problems on GitHub, despite promising a group of mathematicians in August that it would avoid releasing all the results at once. Wired obtained meeting records indicating that OpenAI researcher Sébastien Bubeck threatened mathematician Tristan Buckmaster’s career during a priority dispute over the Navier-Stokes problem, one of the Millennium Prize Problems, which carries a $1 million prize. Mathematician Bryna Kra, a coauthor of the Leiden Declaration on AI in Mathematics, says the company ignored both the community’s direct request and the declaration itself.
In August, OpenAI convened around 40 mathematicians and suggested that its model had solved hundreds of longstanding open problems. According to Northwestern University mathematician Bryna Kra, participants asked the company to publish full papers that researchers could examine and cite, rather than blog posts and tweets. “That request seems to have been ignored,” Kra says in Wired’s investigation.
On October 6, the company announced in a blog post that it had uploaded 722 manuscripts presenting solutions to GitHub; some of the proofs were formalized in Lean, a language for computer verification of proofs. During evaluation, the model attempted around four thousand problems, with each result taking about three hours of computation at the level available through the paid ChatGPT Pro service. For most results, the company did not prepare papers explaining the work.
Wired also disclosed records of an earlier conflict, a priority dispute over the Navier-Stokes problem. New York University mathematician Tristan Buckmaster accused OpenAI of preempting his unpublished work with Anthropic researcher Levent Alpöge. According to those records, Bubeck told Buckmaster, “If you don’t want me to be nice, I don’t have to be nice.” He attributed the urgency to competition with Anthropic: “What would stop Anthropic from devoting all its compute to solving a Millennium Prize Problem?”
Nestor Guillen, a visiting professor at New York University, describes his colleagues’ reaction: “There is a sense of thuggish behavior” by AI companies. He says mathematicians’ concerns center on AI companies and “the concentration of power in one place,” rather than on artificial intelligence itself. OpenAI disputes this characterization and denies making the promise; its spokesperson, Lindsey McCallum, said she was unaware of it.
The Leiden Declaration on AI in Mathematics has been in place since spring 2026. It offers recommendations for researchers, universities, and industry, was coauthored by Kra, and received the International Mathematical Union’s endorsement in June; it has more than four thousand signatures. After the same priority dispute, in which Alpöge claimed that Bubeck offered him the Millennium Prize in exchange for being left off the paper, OpenAI set up its own council of mathematicians, while the community launched the verification tools Hexagon and Palomar. Kra says the declaration, these tools, and the company’s own council have all failed to change OpenAI’s behavior.
According to Wired, some OpenAI employees consider the fate of mathematics as a human profession already settled. Bubeck takes the opposite view: “I see this as an opportunity to expand what mathematicians can do and the impact of their work.”
Wired reports that competition among companies ahead of their IPOs is also accelerating publication; mathematicians are left to verify the results and establish attribution, with no formal means of stopping a release. Kra describes her mixed feelings:
“This is a frightening time... but it is also a truly exciting time.”