OpenAI seats math advisory group after Navier–Stokes claim

On Sept. 21, 2026, OpenAI announced an independent Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study. The move follows OpenAI’s claim that an internal model trained since Aug. 28 has resolved the Navier–Stokes Millennium Prize problem and more than 100 long-standing open problems across most areas of mathematics.

A frontier lab says its model is clearing classic open problems faster than its own mathematicians expected — and is building an unpaid outside brain trust to help review, communicate, and set academic standards around those results. It is advisory governance, not a statute.

The rule in English

Capability claim (OpenAI). Training of a new internal model began Aug. 28, 2026. OpenAI says the model has resolved Navier–Stokes (Clay Millennium Prize) and 100+ open problems; internal mathematicians were surprised by the pace. The company says it is working through responsible broader deployment of math-related AI capabilities.

Community pressure. OpenAI cites mathematicians’ open letter “A Severe Misalignment of AI in Mathematics,” which criticizes using open-problem solving as a benchmark for AI systems and highlights negative externalities for the field.

Advisory group remit. Advise on review and communication of emerging results; help assess significance; advise on dissemination coordination; advise on academic and professional standards of mathematical research; advise on how OpenAI tools can support research and learning.

Independence terms. Operates independently of OpenAI; may offer unsolicited advice; may comment on OpenAI’s impact on mathematics; may make advice public; members not paid by OpenAI; membership can change as the group sees fit. Explicit carve-out: the group will not advise OpenAI on how to pace internal progress on mathematics.

Initial members (IAS-hosted list). François Charles (ENS-PSL); Camillo De Lellis (IAS, GSSI); Timothy Gowers (Collège de France, Cambridge); Martin Hairer (EPFL, Imperial); Nikhil Srivastava (Berkeley, Simons Institute); Ulrike Tillmann (Oxford, INI); Ravi Vakil (Stanford); Edward Witten (IAS); Melanie Matchett Wood (Harvard).

Who wrote it / who it binds

Writer: OpenAI (company post). Group hosted at IAS. Binds OpenAI only as a voluntary consultation setup. Does not bind journals, prize committees, universities, or other labs. Does not certify the Navier–Stokes or open-problem claims.

Who it exempts

OpenAI’s internal pacing decisions (carved out of the group’s remit). Other AI labs. Formal Clay Mathematics Institute prize processes.

What a founder, parent, or worker should do differently

Lab founders: expect parallel “domain advisory” asks wherever models claim to close canonical open problems — math first, then maybe other sciences.

Mathematicians and journals: treat OpenAI’s resolve claims as company disclosures until independent verification and dissemination standards the advisory group helps shape actually land.

Parents and workers: not a consumer product rule — research-community governance around AI-generated proofs.

What’s next

• First public advisory advice or challenge to an OpenAI math claim.
• Independent verification path for the Navier–Stokes claim (Clay / peer review).
• Whether other labs seat matching unpaid domain boards — or keep results closed.
• Tension between “inform the community” and the carve-out that blocks pacing advice.

Source: OpenAI, “Advisory Group on Mathematics and Artificial Intelligence,” Sept. 21, 2026.