OpenAI has formed a nine-member independent advisory panel of elite mathematicians to help it manage the release of research results generated by its internal models, following a string of botched announcements that damaged its standing in the mathematics community. The group, called the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), was announced Monday and will be hosted by the Institute for Advanced Study (IAS) in Princeton, New Jersey — the former intellectual home of Albert Einstein, John von Neumann, and J. Robert Oppenheimer. Its first assignment: figure out how OpenAI should disclose more than 100 long-standing open problems that the company says its unreleased internal model has resolved.
The panel is stacked with researchers drawn from Stanford, Harvard, Oxford, Cambridge, ETH Zurich, Imperial College London, and Switzerland's EPFL, and includes multiple Fields Medal and MacArthur genius grant recipients. Members will not be paid by OpenAI, have signed no restrictive agreements beyond standard confidentiality, and can change the group's membership as they see fit. OpenAI has committed to giving AGMAI early access to research and the freedom to publish advice the company did not request.
The formation follows what The Verge described as a reputational crisis for OpenAI in the mathematics world, driven by the company's handling of a series of high-profile results including its claim to have resolved the Navier-Stokes Millennium Prize problem. Mathematicians accused the company of scooping active researchers, failing to credit prior human work, and communicating results in ways that fell short of professional standards.
“The group will advise on the review and communication of emerging results: they will help OpenAI assess their significance, advise on how to coordinate their dissemination, and advise on academic and professional standards of mathematical research.”— OpenAI, company statement
Key facts
- 01OpenAI announced a nine-member Advisory Group on Mathematics and Artificial Intelligence (AGMAI) on Monday, hosted by Princeton's Institute for Advanced Study.
- 02OpenAI says its unreleased internal model has resolved more than 100 long-standing open problems, including the Navier-Stokes Millennium Prize problem.
- 03AGMAI members are unpaid, have signed no restrictive agreements beyond confidentiality, and rely on the IAS for technical support.
- 04The panel draws from Stanford, Harvard, Oxford, Cambridge, ETH Zurich, Imperial College London, and EPFL, with multiple Fields Medals among its members.
- 05AGMAI's first stated priority is advising OpenAI on how to coordinate the release of the 100-plus results the company says its model has produced.
AI Chat Daily covered OpenAI's IAS partnership at the initial announcement; the additional detail this week is the composition of the group and the first substantive public statement from a member.
That statement came from Martin Hairer, a professor of mathematics at Imperial College London and EPFL, who published a blog post Tuesday responding to what he called a torrent of misinformation about AGMAI. Hairer wrote that not all nine members had been approached by OpenAI before the group coalesced, and stressed that AGMAI is genuinely independent of OpenAI and other frontier AI labs. He also acknowledged the obvious PR risk: members would be naïve to believe the AI labs won't try to spin whatever the group says.
Hairer framed the immediate problem in blunt terms. Refusing to engage with AI labs on principle, he wrote, is a hard ostrich approach that helps no one, given how deeply AI has already reshaped mathematical research and questions of attribution, priority, and human understanding.
“It is a fact that AI companies have in recent months been producing some high profile mathematical results and that their publication and dissemination has been falling far short of acceptable mathematical practice, whichever way that is defined.”— Martin Hairer, mathematics professor at Imperial College London and EPFL, AGMAI member
Not everyone in the field is convinced the panel is representative. Francesco Fournier-Facio, an incoming mathematics professor at Heriot-Watt University in Scotland, told The Verge that greater engagement with mathematicians can only make things better but described AGMAI as feeling like an ivory tower. Simon Machado, a researcher at ETH Zurich, said the whole thing seemed shrouded in mystery and questioned how the nine members were selected — the group's website says only that it came together after OpenAI approached some of its members.
Machado's deeper concern is representation. The panel's members occupy elite academic positions insulated from the labor-market and attribution pressures now hitting working mathematicians, whose research programs and career prospects are being reshaped by AI systems producing rapid-fire results.
“They're amazing mathematicians, but I don't know if they are the people I want to represent me in more political questions”— Simon Machado, researcher at ETH Zurich
The group has not yet decided what specific advice it will give. For now it is gathering input through colleague conversations, a public feedback form, and discussion on Proofs and Prompts, a blog that has become a central forum for mathematicians reckoning with AI's impact on the field. AGMAI's website lists coordinating the disclosure of OpenAI's backlog of results — the 100-plus resolved problems and a second Millennium Prize problem the company says it is closing in on — as its explicit first priority.
OpenAI has structural incentives to want this to work. The company's math announcements over the past few months generated exactly the kind of scientific-credibility backlash that undercuts the marketing value of a breakthrough. An independent panel with Fields Medalists on it gives OpenAI a route to publish future results with peer legitimacy pre-baked into the rollout, while giving working mathematicians a channel to challenge attribution and communication practices before they hit the news cycle. Whether AGMAI functions as a genuine check or as reputational cover will be decided by the first substantive disagreement between the group and the company — and by whether OpenAI publishes AGMAI's dissenting advice when that happens.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
