OpenAI’s new math advisory group faces confusion over independence after rollout missteps

OpenAI’s attempt to mend strained relations with the mathematical community by forming an independent advisory group has instead sparked confusion about the group’s independence. The Advisory Group on Mathematics and Artificial Intelligence (AGMAI) was announced on September 21, 2026, in a guest post on mathematician Terence Tao’s blog. The group consists of nine elite mathematicians and was described as operating independently from OpenAI, with members free to challenge the company publicly and publish their advice.

However, OpenAI’s prominent rollout of the announcement led many to assume AGMAI was an OpenAI-appointed body. This confusion stemmed partly from AGMAI’s website stating it formed after OpenAI approached some of its eventual members about establishing an advisory board, though it later decided to proceed independently. Martin Hairer, a member of AGMAI and a mathematics professor at Imperial College London and EPFL, emphasized the group’s independence, stating it receives no financial, technical, or other support from OpenAI and that members are not restricted from speaking publicly beyond standard confidentiality requirements for advanced research access.

The episode follows a pattern of OpenAI’s math-related announcements drawing criticism. Earlier this year, OpenAI’s solution to a Millennium Prize problem sparked disputes over credit and treatment of mathematicians, further straining trust between the company and the math community. Hairer acknowledged the risk of AI companies attempting to spin the group’s involvement to their advantage, noting that OpenAI’s handling of past announcements had heightened community suspicion.

AGMAI’s formation comes as OpenAI says it has produced scores of additional mathematical results with an unreleased model, a prospect that has already prompted concern among researchers about the potential impact on their field. The group’s role includes advising on the review and communication of these emerging results, though OpenAI has stated the group will not advise on internal pacing of progress.