Advisory Group on Mathematics and Artificial Intelligence announced

The advisory group is a concrete response to AI‑generated mathematics

The Advisory Group on Mathematics and Artificial Intelligence (AGMAI) was announced on 21 September 2026 as an independent, unpaid panel of leading mathematicians hosted by the Institute for Advanced Study. Its immediate task is to advise OpenAI on how to coordinate the public release of a large batch of significant mathematical results that OpenAI’s internal model claims to have produced.

“We are currently facing the very specific challenge of advising OpenAI on how to coordinate the release of a large number of significant results in mathematics that they report have been produced by their internal model.” – AGMAI announcement

The group’s purpose is to serve as a communication channel between the mathematical community and AI companies, ensuring that any dissemination of AI‑generated results respects academic standards, transparency, and the long‑term health of mathematics.

Membership and governance

The panel consists of nine eminent researchers:

  • François Charles (ENS‑PSL)
  • Camillo De Lellis (IAS, GSSI)
  • Timothy Gowers (Collège de France, Cambridge)
  • Martin Hairer (EPFL, Imperial College London)
  • Nikhil Srivastava (Berkeley, Simons Institute)
  • Ulrike Tillmann (Oxford, INI)
  • Ravi Vakil (Stanford)
  • Edward Witten (IAS)
  • Melanie Matchett Wood (Harvard)

The members do not receive payment, and the group operates independently of any AI firm. It will publish its recommendations on the public website agmai.org.

Core concerns raised by the community

The announcement sparked a vigorous discussion on Hacker News and the blog comments. Several recurring themes emerged:

1. Transparency vs. gate‑keeping

“The only thing I want to see is the problem statements, solutions, and associated Lean proofs. Anything else is gatekeeping.” – anonymous commenter

Many participants argued that the community should have immediate, unrestricted access to the problem statements and proofs, fearing that a small advisory board could become a new bottleneck for knowledge dissemination.

2. Trust in OpenAI’s motives

“OpenAI is using mathematics as a PR tool. This devalues math and mathematicians.” – gpt5 comment

Skepticism about OpenAI’s intentions was widespread. Commenters noted the company’s history of crisis management and questioned whether the advisory group could meaningfully influence OpenAI’s business decisions.

3. Academic independence and legitimacy

“Is it a good idea to help their crisis management? It is unrealistic to think this group can change the way OpenAI does business.” – Burt Totaro

Some mathematicians worried that collaborating with a profit‑driven firm might compromise the perceived independence of the discipline.

4. Practical recommendations for release

Several concrete ideas were proposed:

  • Publish problem statements and hash‑verified proofs first, with full write‑ups following after a short embargo (e.g., one month). – anonymous comment
  • Create a public, searchable repository independent of arXiv for AI‑generated results, allowing community‑wide verification and annotation. – Grigori Avramidi
  • Provide compute grants to external researchers so they can explore, verify, and improve the AI‑generated proofs. – anonymous comment

5. Impact on careers and education

Early‑career researchers expressed anxiety about job security and the relevance of traditional PhD training:

“If I spent my entire salary on tokens, I could get more results than a human mathematician.” – anonymous PhD student

Others called for a re‑evaluation of curricula to incorporate AI tools while preserving the ability to solve problems without assistance.

Divergent viewpoints

The discussion was not monolithic. Some participants praised the rapid, coordinated response of the mathematical community:

“The math community has come together, assessed the situation, and acted calmly and rationally.” – @ipdashc (Hacker News)

Others warned that the advisory board could become an oligarchic gate‑keeper, concentrating control over the release of breakthroughs in the hands of a few elite mathematicians.

What the advisory group can realistically achieve

Given its stated remit—advising on presentation and release of results, not on the pace of AI research—the group can:

  1. Define standards for disclosure (e.g., require Lean certificates, hash values, and clear provenance).
  2. Recommend a release schedule that balances priority claims with community access.
  3. Facilitate a transparent review process by publishing recommendations and rationales on agmai.org.
  4. Advise on ethical considerations, such as withholding cryptographically sensitive results.

What it cannot do is halt OpenAI’s internal research or dictate corporate strategy.

Outlook

The formation of AGMAI marks a historic moment: the first formal, high‑level interface between the world‑leading AI labs and the global mathematics community. The next weeks will reveal whether the group can steer OpenAI toward open, verifiable, and community‑beneficial practices, or whether it will become a symbolic body with limited influence.

Key take‑aways for readers:

  • The advisory group is independent, unpaid, and will publish its recommendations publicly.
  • Immediate community concerns focus on transparency, avoiding new gate‑keeping, and ensuring that AI‑generated results are accessible for verification.
  • Practical proposals include hash‑verified proof releases, a dedicated repository, and compute grants for external researchers.
  • The group’s impact will depend on how OpenAI responds to its advice and how the broader mathematical community engages with the forthcoming AI‑generated results.

Sources

Related