The OpenAI math advisory group was announced on Monday 21 September 2026 with nine mathematicians as founding members. The Advisory Group on Mathematics and Artificial Intelligence is hosted at the Institute for Advanced Study in Princeton. OpenAI says it will not advise on how fast the company pushes its mathematics research.
Who Is on the OpenAI Math Advisory Group
Nine mathematicians were named as initial members. OpenAI has described the group as independent, and its members take no payment for the work.
- François Charles
- Camillo De Lellis, of the Institute for Advanced Study
- Timothy Gowers
- Martin Hairer
- Nikhil Srivastava
- Ulrike Tillmann
- Ravi Vakil
- Edward Witten
- Melanie Matchett Wood
The group is not confined to OpenAI. It can advise other artificial-intelligence companies as well, and it intends to publish its recommendations rather than deliver them privately. OpenAI has not said how often those recommendations will appear or where.
What the Group Is Allowed to Advise On
The remit is about how results are judged and released, not about what OpenAI builds. According to OpenAI’s announcement, the group will:
- Review and communication: advise on the review and communication of emerging mathematical results.
- Significance: help OpenAI assess how important a given result actually is.
- Dissemination: advise on how to coordinate the release of results.
- Standards: advise on the academic and professional standards of mathematical research.
- Tools: advise on how OpenAI’s tools can support mathematical research and learning.
The One Thing the Group Cannot Touch
OpenAI drew the boundary in its own announcement: the group “will not be responsible for advising us on how to pace our internal progress on mathematics.” That sentence removes from the group’s reach the question the mathematical community has been loudest about.
The Institute for Advanced Study set out the same limit from its side, stating that although the group will give advice, “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.” Neither OpenAI nor the Institute has described any mechanism by which the group’s advice would bind anyone.
The 100-Problem Claim Behind the Announcement
The advisory group arrives attached to a claim. OpenAI says an internal model has resolved more than 100 long-standing open problems across most areas of mathematics. The company has not named the model, published the list of problems, or said whether the proofs have been checked by mathematicians outside OpenAI.
That figure follows the claimed solution to the Navier-Stokes Millennium Prize problem that OpenAI published on 8 September 2026. Our earlier report explains why the OpenAI Navier-Stokes proof faces a two-year wait under the Clay Mathematics Institute’s own rules before any prize could be awarded. The 100-problem claim has had no comparable public verification process attached to it.
What 27 Fields Medallists Said Ten Days Earlier
The group’s announcement lands after a public rebuke from the top of the field. On 11 September 2026, a declaration titled “A Severe Misalignment of AI in Mathematics” was published at mathandai.org. As of 22 September 2026 it carries the signatures of 27 Fields Medal recipients, from winners of the 1974 medal through to 2026; TechCrunch counted 25 signatories when it reported on the advisory group on 21 September.
The declaration’s argument is about purpose rather than accuracy. Its central line is that “solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight”, and that forgetting this “may turn the tool against the primary goal.” It does not ask AI companies to stop. It says the issues “must be addressed urgently, in the mathematical community, by the companies developing these technologies and, more broadly, by a society that will confront similar problems.”
TechCrunch reported that of the nine advisory group members, only Camillo De Lellis signed that declaration. The wider argument about whether frontier labs should slow down is covered in our report on the “pace the frontier” plan and its rejection.
What Is Still Unknown
As of 22 September 2026, several basic questions about the arrangement have no published answer.
- The results: OpenAI has not released the 100-plus problems, the proofs, or the identity of the model that produced them.
- Verification: no independent review process for those results has been announced.
- Authorship: neither OpenAI nor the group has said how AI-generated proofs should be credited or cited.
- Effect: the group has issued no recommendations yet, and none of its advice has been tested against a decision OpenAI did not want to make.
Frequently Asked Questions
What Is the OpenAI Math Advisory Group?
It is the Advisory Group on Mathematics and Artificial Intelligence, announced by OpenAI on 21 September 2026 and hosted at the Institute for Advanced Study in Princeton. Nine mathematicians serve as its founding members, without payment.
Can the Group Stop or Slow OpenAI’s Research?
No. OpenAI stated that the group will not advise it on how to pace its internal progress in mathematics, and the Institute for Advanced Study said the group has no decision-making power at any AI company.
Does the Group Work Only for OpenAI?
No. The group is able to advise other artificial-intelligence companies as well, and it plans to publish its recommendations rather than keep them internal.
Have the 100 Solved Problems Been Verified?
Not publicly. OpenAI has said an internal model resolved more than 100 long-standing open problems but has not released the problems, the proofs, or details of any outside review.
Why Are Mathematicians Objecting?
A declaration published on 11 September 2026 and signed by 27 Fields Medal recipients argues that treating problem-solving as the goal mistakes a proxy for the real aim of mathematics, which it describes as conceptual understanding and insight.
