Press "Enter" to skip to content

OpenAI Forms Elite Advisory Panel to Guide Release of AI‑Generated Math Breakthroughs

OpenAI announced on Monday the creation of an independent advisory body, the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), composed of nine highly distinguished mathematicians. The panel is intended to help the company and the broader AI industry navigate how emerging mathematical results are evaluated, communicated and released.

Panel composition and mandate

The nine members are drawn from institutions that include Stanford, Harvard, Oxford and Cambridge, and collectively hold multiple Fields Medals and MacArthur “genius” grants. The group will be hosted by the Institute for Advanced Study (IAS) in Princeton, New Jersey, a historic center of mathematical research that once counted Albert Einstein, John von Neumann and J. Robert Oppenheimer among its affiliates.

OpenAI describes AGMAI as operating independently, with the freedom to offer unsolicited advice, publicly comment on the impact of AI on mathematics, and make its recommendations public. Members will receive early access to OpenAI’s research, but they will not be compensated financially and the panel may change its membership at its own discretion. The company said the group’s value depends on members being able to exercise independent judgement and to challenge OpenAI’s own decisions.

The advisory group’s stated responsibilities include assessing the significance of new results, coordinating their dissemination, advising on academic and professional standards for mathematical research, and suggesting ways OpenAI’s tools can support mathematical inquiry and learning.

Origins, independence and community reaction

According to a blog post by panel member Martin Hairer, the group formed after OpenAI approached several of the mathematicians about establishing an external advisory board. After discussions, the scholars chose to launch an independent organization and to invite additional members. Hairer emphasized that not all members were initially contacted by OpenAI and that the group is “genuinely independent of OpenAI and any other of the so‑called ‘frontier’ AI labs.” He added that, aside from confidentiality requirements, members have signed no restrictive agreements and receive no payment, relying on the IAS for technical support.

The announcement surprised many researchers. Simon Machado of ETH Zurich described the move as “shrouded in mystery” and questioned how the nine members were selected. Francesco Fournier‑Facio, an incoming professor at Heriot‑Watt University, called the panel “an ivory tower” but acknowledged that greater engagement with mathematicians could be beneficial. Both scholars expressed uncertainty about whether the small, elite group can represent the broader mathematical community or address more political concerns affecting typical researchers.

Context of recent AI‑produced mathematics

OpenAI’s decision follows a series of high‑profile announcements in which its internal model claimed to resolve the Navier‑Stokes Millennium Prize problem and more than 100 long‑standing open problems across diverse areas of mathematics. The company has also indicated progress toward another Millennium Prize problem. Critics have accused OpenAI of scooping researchers, failing to credit human contributions and, in some cases, appropriating work generated with its models.

AGMAI’s first priority, as stated on its website, is to advise OpenAI on how to coordinate the release of the large number of significant mathematical results the company reports it has produced. Hairer noted that AI companies have been generating “high profile mathematical results” whose publication practices fall short of accepted mathematical standards, and that the panel will gather input through conversations with colleagues, a public feedback form and discussion on the Proofs and Prompts blog.

Whether the advisory group can earn the trust of the wider mathematics community will shape its effectiveness in mitigating the reputational fallout from OpenAI’s recent missteps and in guiding future disclosures of AI‑generated breakthroughs.