OpenAI's 722 math manuscripts spark mathematician backlash
According to a 36Kr article, OpenAI published 722 math manuscripts on GitHub on Oct. 6, 2026—covering topics such as the Kakeya conjecture, a quasi-Riemann hypothesis and special cases of the Hodge conjecture—released as 372 results, and it also said its internal model found a counterexample to the Navier-Stokes problem, prompting a credit dispute with NYU mathematician Tristan Buckmaster. The piece says 25 Fields medalists signed a protest and nearly 7,000 mathematicians signed a petition, with Terence Tao's Human Mathematics Association urging researchers to stop collaborating with OpenAI, while only about 42% of 719 documents had Lean formal verification and three manuscripts were retracted the next day. The allegations have not been independently verified and OpenAI denies them.
It puts verification, authorship and collaboration norms for AI-generated mathematics in the spotlight, potentially reshaping how AI labs work with academia.