OpenAI has established an independent advisory group of elite mathematicians to manage communication of its advanced AI-generated mathematical results, as the company prepares to release what it describes as "scores more" breakthroughs.
The Advisory Group on Mathematics and Artificial Intelligence (AGMAI) was announced Sept. 21, tasked with advising OpenAI and other frontier AI laboratories on reviewing and communicating emerging results. The group consists of nine members operating independently from OpenAI, announced via a guest post on UCLA mathematics professor Terence Tao's blog. Tao, a Fields Medalist, has been a vocal critic of how AI companies handle mathematics research.
OpenAI emphasized AGMAI's independence in its own announcement, stating members are free to challenge the company publicly, publish their advice, and offer unsolicited guidance. The company did specify limits, however, noting the group "will not be responsible for advising us on how to pace our internal progress on mathematics."
The rollout created confusion. OpenAI's public announcement was substantially larger than the independent disclosure, leading many mathematicians to initially assume AGMAI was an OpenAI-appointed body—a reasonable assumption given that AGMAI's website indicates OpenAI approached some eventual members about forming the group.
OPENAI'S IMMEDIATE CHALLENGE: releasing what one AGMAI member called a "tidal wave of breakthroughs" from an unreleased model. The prospect is already stirring concern among researchers over its potential impact on the field.
An AGMAI member, Hairer, said the group is gathering input through conversations with colleagues, a public feedback form, and discussions on Proofs and Prompts, a blog serving as a key forum for mathematicians grappling with AI's impact. Hairer described the process of creating the group as "messy and confusing."
The real business risk for OpenAI is operational: integrating advanced mathematical capabilities into its product roadmap without alienating the academic community that supplies both talent and credibility. In an AI sector where proximity to frontier research drives hiring and partnership deals, community relations are not peripheral—they are a competitive input. OpenAI's ability to generate significant mathematical advancements suggests genuine capability. Consistent mismanagement of how it shares that work is a friction point that could matter.
