OpenAI forms independent advisory group after saying AI resolved 100-plus math problems

OpenAI said on 21 September 2026 that an internal model it began training on 28 August has produced a solution to the Navier–Stokes Millennium Prize problem and has since resolved "more than 100 long-standing open problems across most areas of mathematics," according to the company's own blog post. TechCrunch reported the announcement the same day and said OpenAI is forming an independent group of mathematicians to review how such results are handled.

What was announced

• The new Advisory Group on Mathematics and Artificial Intelligence is hosted at the Institute for Advanced Study (IAS) in Princeton, New Jersey, TechCrunch reported.
• Nine mathematicians were named as initial members, including Timothy Gowers, Martin Hairer, Edward Witten and Melanie Matchett Wood, according to OpenAI's blog post. TechCrunch noted that only one member, IAS's Camillo De Lellis, had also signed a mathematicians' open letter criticising the AI industry's pace.
• The group "will serve as a bridge to the mathematical community and broader public, giving mathematicians a voice in how we move forward," OpenAI said, but it "will not be responsible for advising us on how to pace our internal progress on mathematics."
• The Institute for Advanced Study said in its own statement, quoted by TechCrunch: "we do not have decision making power at any AI company, and the responsibility for the decisions made by any company will rest with that company."
• The move follows an open letter, "A Severe Misalignment of AI in Mathematics," which TechCrunch said was signed by 25 Fields Medal-winning mathematicians arguing that AI labs are "threatening their intellectual work" by racing to solve famous problems as benchmarks.
• OpenAI published the proof on its own site together with a formalisation in the Lean proof assistant; no independent mathematical review of the result has been published, and TechCrunch described the 100-plus figure as OpenAI's claim.

Why it matters

A leading AI lab now says its own model can resolve a problem that OpenAI's accompanying post says has stood unresolved for roughly 90 years, and it is answering academic pushback with an advisory body that reviews disclosure, not development pace — a distinction that will shape how far self-regulation goes as frontier labs race on scientific benchmarks. For regulators and universities outside the United States, including in Europe and the Caspian region, the episode underscores that oversight of high-impact AI research results is currently voluntary and company-designed, with no equivalent binding review mechanism yet in force under instruments such as the EU AI Act.