CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 818 -0.1%
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 711 -0.1%
CL-OP5X 709 -0.1%
KIMI-K3X 699 -0.1%
CL-OP46H 694 -0.1%
CL-FAB5H 692 -0.1%
CL-OP55H 692 -0.1%
CL-OP47H 683 -0.1%
GEM-37FH 648 -0.2%
GEM-38FH 633 -0.2%
INKL 531 —
CL-OP47 529 -0.2%
GPT-56S 529 -5.9%
GEM-3P 493 —
CL-OP46 488 —
CL-OP48 481 —
GEM-31P 453 -0.2%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 885 —
GPT-56T 861 —
GPT-6S 825 —
CL-OP55X 818 -0.1%
MUSE-SPK 793 —
QWEN-38X 779 —
GPT-6A 778 —
CL-OP5H 711 -0.1%
CL-OP5X 709 -0.1%
KIMI-K3X 699 -0.1%
CL-OP46H 694 -0.1%
CL-FAB5H 692 -0.1%
CL-OP55H 692 -0.1%
CL-OP47H 683 -0.1%
GEM-37FH 648 -0.2%
GEM-38FH 633 -0.2%
INKL 531 —
CL-OP47 529 -0.2%
GPT-56S 529 -5.9%
GEM-3P 493 —
CL-OP46 488 —
CL-OP48 481 —
GEM-31P 453 -0.2%
GPT-6L 436 —
← Back to feed

OpenAI Forms External Math Advisory Group to Validate Frontier AI Results

OpenAI announced an Advisory Group on Mathematics and Artificial Intelligence on September 21. The group will sit outside OpenAI’s research organization and advise on three specific functions: reviewing and communicating emerging results, coordinating their dissemination, and upholding academic and professional standards for mathematical research.

What the Group Does

The advisory group has a narrow and specific mandate:

  • Significance assessment: Help OpenAI determine whether an AI-produced mathematical result is genuinely novel and important, or an artifact of training data or benchmark contamination.
  • Dissemination coordination: Advise on how and when to publish results, including preprint strategy, peer review, and communication to the broader math community.
  • Standards setting: Advise on what constitutes valid mathematical research when the result is produced or assisted by an AI system.

The group will also advise on how OpenAI’s tools can support mathematical practice, a mandate that covers both research acceleration and pedagogy.

Context

The timing is not incidental. The past six months have seen escalating tension between AI labs and professional mathematicians over how AI-produced results are characterized.

In September 2026, 25 Fields Medalists signed an open letter accusing AI labs of severe misalignment with mathematical standards. Earlier, the Leiden Declaration (June 2026) launched with roughly 130 initial signatories and grew to more than 4,000, all arguing that AI is threatening the integrity of mathematical proof by flooding the literature with claims that cannot be independently verified.

OpenAI has simultaneously been advancing its most public mathematical claims, including reports of solving the Navier-Stokes Millennium Prize problem and over 100 other open problems, attributed to an unreleased internal model described as significantly more capable than GPT-6 Astra. The gap between those claims and the mathematics community’s willingness to credit them has been a recurring source of friction.

An external advisory group with credentialed mathematicians gives OpenAI a mechanism to pre-screen results before announcement and a credible imprimatur when publishing. For the math community, it creates a point of contact without requiring wholesale endorsement of AI-produced work.

What It Does Not Do

The group is advisory. It does not peer-review OpenAI’s work in the formal sense, does not have access to model weights or training data, and has no enforcement authority over what OpenAI publishes. OpenAI disclosed the independence terms in its announcement: the group is hosted externally at the Institute for Advanced Study, members are unpaid, can offer unrequested public advice, and can alter membership independently.

OpenAI named nine initial members: Francois Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood.