OpenAI published 722 mathematics manuscripts at once on October 6, 2026, in a public GitHub repository, covering 372 problems across fields that include geometry, algebra and computer science. The company credited the results to an internal AI model and called the release an effort "to push the frontier of human knowledge and enable further progress in mathematics," Retraction Watch reported.
A day later it pulled three of them. A single wrong sign had broken the reasoning in one paper, and two other papers built on that reasoning fell with it. The withdrawn papers include "The rational Hodge conjecture for products of K3 surfaces." OpenAI said it had also "revised 14 other manuscripts with proof repairs, corrected statements, clearer hypotheses and dependencies, and one correction to an obsolete citation."
The repository now lists 719 manuscripts in 372 families and says about 42% of its top-line results have been formalized in Lean, an open-source programming language that checks math proofs. An OpenAI spokesperson said roughly 50% of the results were released unconfirmed, and that the company's collaborators at the Advisory Group on Mathematics and Artificial Intelligence had recommended releasing them without waiting for "full formalization."
"We welcome scrutiny and feedback from the mathematical community," the spokesperson told Retraction Watch, adding that the errors were found during an audit. Dan Roberts, a research lead at OpenAI, wrote on X that the company will "continue to update the repo with new formalizations and with any errata we notice."
Alex Townsend, an associate professor of mathematics at Cornell University, said OpenAI "should have announced the manuscripts that were lean verified first." Andrew Sutherland, a senior research scientist in the mathematics department at the Massachusetts Institute of Technology, called the quick withdrawal "the responsible thing to do," but added that "it will take a lot more than that to earn back the trust they have lost."
