Mathematicians to advise OpenAI on releasing results
On 21 September 2026 OpenAI announced it would work with an independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study. Nine mathematicians, among them Timothy Gowers, Martin Hairer and Edward Witten, serve without pay and may publish advice nobody asked for. Their first task is to advise on how to coordinate the release of a large number of results that, according to OpenAI, its internal model has produced.
Why it matters
Ten days after the Fields medallists' declaration, the lab whose pursuit of open problems prompted it handed the question of how results are released to mathematicians who do not depend on it. The group itself writes that OpenAI proposed an external board and that they instead created an independent one, so the frame was set by the mathematicians, not the company.
Members: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, Melanie Matchett Wood. The group will advise on verifying and releasing new results and on academic standards but, according to OpenAI, not on the pace of its internal progress. It promises to publish its recommendations on its own site and offers them to any company whose models are likely to have a significant impact on mathematics. Ten days earlier, on 11 September, the Clay Mathematics Institute wrote that the Navier-Stokes problem had "apparently been settled" and that the evaluation under the prize rules is deliberately unhurried. What the record does not claim: that OpenAI's internal model, which the company began training on 28 August, has really solved more than a hundred open problems; that is the company's statement. The group had published no recommendations as of 28 September 2026.