On Monday OpenAI revealed the creation of an independent panel of mathematicians designed to advise the company and other AI labs on how to interact with mathematical research and the broader mathematics community. The panel will be hosted by the historic Institute for Advanced Study (IAS) in Princeton, New Jersey, a venue once occupied by Albert Einstein, John von Neumann and J. Robert Oppenheimer. The new body, officially called the Advisory Group on Mathematics and artificial intelligence (AGMAI), is intended to serve as a bridge between cutting-edge AI systems and the established norms of mathematical publishing.
The announcement follows a series of controversial AI-generated announcements that many in the field regard as a public-relations mishap. In recent months OpenAI claimed that an internal, unreleased model had solved the Navier-Stokes Millennium Prize problem and more than one hundred other long-standing open problems across diverse areas of mathematics. Critics accused the company of scooping active researchers, neglecting to credit prior human work, and presenting results in a manner that diverged from standard peer-review practices. The backlash highlighted a gap between the speed of AI-driven discovery and the slower, more deliberative culture of the mathematical community.
The nine-member AGMAI comprises mathematicians drawn from Stanford, Harvard, Oxford, Cambridge, ETH Zurich, Imperial College London and the Swiss Federal Institute of Technology Lausanne (EPFL). Among them are multiple Fields Medal recipients and scholars who have earned MacArthur “genius” grants. Members will not receive financial compensation from OpenAI and have signed only standard confidentiality agreements; no restrictive clauses limit their public commentary. The group retains the right to modify its roster at any time, and it will be granted early access to the company’s research in order to fulfill its advisory duties.
According to statements from OpenAI AGMAI will operate independently with the freedom to issue unsolicited advice, comment publicly on the impact of AI on mathematics, and make its recommendations openly available. Its charter includes evaluating the significance of emerging results, advising on coordinated dissemination, and ensuring that the release of findings adheres to accepted academic and professional standards. The panel is also tasked with exploring how AI tools can support mathematical learning and research, and with proposing best-practice guidelines for future collaborations between AI labs and mathematicians.
Reactions from the community have been mixed. Simon Machado of ETH Zurich described the formation as “shrouded in mystery” and questioned whether a nine-person elite board can truly represent the diverse interests of the global mathematics community. Francesco Fournier-Facio, an incoming professor at Heriot-Watt University, called the initiative “an ivory tower” and expressed ambivalence about its capacity to address the concerns of ordinary researchers. In a blog post published Tuesday, AGMAI member Martin Hairer emphasized that the group is “genuinely independent of OpenAI and any other of the so-called ‘frontier’ AI labs,” and reiterated the core principle: “The other fundamental fact is that our overarching aim is to represent as well as we possibly can the interests of the mathematical community.” Hairer also warned against an “ostrich” attitude that would ignore AI’s growing influence.
The advisory board is currently gathering input through informal conversations with colleagues, a public feedback form, and discussions on the Proofs and Prompts blog, which has become a focal point for mathematicians grappling with AI-driven breakthroughs. Its first concrete task is to guide OpenAI on how to coordinate the public release of the more than one hundred results the company claims its internal model has produced. While the panel’s independence is emphasized, skeptics note that the ultimate impact will depend on whether the company is willing to follow the group’s recommendations. Trust from the broader mathematical community will be the decisive factor in turning the advisory effort into an effective safeguard against future publishing missteps.
