OpenAI is facing a new issue: its efforts to correct its past mistakes keep producing fresh ones. The firm has recently made notable advances in mathematics, followed by awkward public statements about them. Now it has added another layer to the trouble. It attempted to establish an independent advisory group to supervise future announcements, and that approach has turned into a separate problem of its own.
AGMAI stands for the Advisory Group on Mathematics and Artificial Intelligence. It was introduced on September 21st through a guest post on the blog of UCLA mathematics professor Terence Tao, a Fields Medalist who has openly criticized how AI companies behave toward mathematics. The purpose of the group is to serve as an independent body of nine elite mathematicians offering guidance to OpenAI and other frontier AI labs on “the review and communication of emerging results.”.
How the Group Came Together
Once the announcement came, doubts about independence followed at once. A great many mathematicians, and almost all those outside the community, first heard of AGMAI through OpenAI’s far more prominent rollout. That caused widespread confusion over whether the group acted on its own or stood as a body appointed by OpenAI.
According to the group’s own website, it came into existence when OpenAI reached out to some of its eventual members about starting an advisory board. Instead of joining that board, those members chose to break away and bring other people on board themselves. It is not known which of the nine members OpenAI first contacted.
Hairer on Independence
Martin Hairer, who teaches mathematics at Imperial College London and the Swiss Federal Institute of Technology in Lausanne (EPFL), is part of AGMAI. He made it clear that the group operates separately from OpenAI. Hairer gave these remarks during a video call from a lecture hall.
The group receives no financial, technical, or other support from the company, Hairer said. None of its members have signed agreements limiting what they can say or do beyond standard confidentiality requirements needed for advanced access to research.
He described the preceding few days as a “very frustrating” and “intense” experience. “We don’t work for OpenAI, are not paid by them, and it’s totally independent,” he said. The group is also open to working with other frontier AI labs, though he declined to say which.
Why the Announcement Failed
The announcement by OpenAI provoked Hairer’s exasperation, and he addressed it in a blog post posted during the uproar. He admitted that AI firms would probably attempt to twist how the group’s involvement to their advantage. Given the group had emerged partly in light of OpenAI’ handled past mathematical breakthroughs, with that risk sitting at the forefront.
The rapid descent of OpenAI’s solution to a prestigious Millennium Prize problem into disputes over credit, scooping, and its treatment of mathematicians made the community suspicious.
OpenAI’s announcement did nothing to aid the young group in establishing its independence, according to Hairer. AGMAI must earn the trust of mathematicians so they can help those researchers manage their increasingly difficult relationships with AI companies.
“The way they phrased their announcement didn’t really help,” Hairer said. “If you read it exactly in detail, there’s nothing wrong in what they say,” he noted, adding that the company is “careful” in its wording. “They obviously are very good PR people,” he said, laughing.
The Group’s Urgency Problem
Hairer laughed and said, “So we, well, you know, started two days ago. So it’s not like we have a big master plan.”, after which the group’s formation, its communications with the mathematical community — feels as if it came together in a hurry. Asked last Wednesday about AGMAI’s plans were announced. That sequence marked the entire endeavor.
The rush may stem from OpenAI holding back a batch of findings it wants to put out as quickly as it can. After solving the Navier-Stokes Millennium Prize problem, the company says its unpublished internal model has since solved more than 100 long-standing open questions across several branches of mathematics.
The Trust Gap
The irony is that OpenAI’s announcement process may have undermined the very trust it was trying to rebuild. Mathematicians need to believe AGMAI can speak freely and honestly about OpenAI’s work. A group that looks appointed, or whose members look restricted in what they can say, will not carry that weight.
The ability for OpenAI’s public statements stress the group’ to operate independently and for its staff to publicly question the company and release their guidance is real. Still, how the public perceives control matters just as much as the actual legal situation. Even when a team eventually separates from its corporate roots, the mark of where it began lingers.
What OpenAI Needs to Do Differently
OpenAI tends to act with haste. It moves swiftly, makes announcements quickly, and afterwards handles the consequences. This sequence has played out through several mathematical breakthroughs, giving rise to an impression that the company regards mathematics more as a trophy cabinet than as a genuine field of study.
The company should hold back. It should allow AGMAI time to establish its own identity and connections within the mathematical community before releasing any results. The company needs to trust the group to manage communications on its own, without being told what to say.
The habit of treating every announcement as a triumph must end. OpenAI has resolved genuine issues, yet its manner of speaking about them has left mathematicians watching from the sidelines instead of working alongside the company. It is a loss for all parties.
A Comparison of OpenAI’s Announcements
| Announcement | Result | Reception |
|---|---|---|
| Navier-Stokes Millennium Prize problem | Breakthrough confirmed | Credit disputes, scooping fights |
| 100+ open problems | Unreleased model claims | Dread among researchers |
| AGMAI formation | Independent advisory group | Confusion over independence |
A steady pattern emerges from the comparison: OpenAI draws attention, then devotes weeks or months to fixing whatever trouble follows. That sequence holds true for AGMAI as well. The group was formed to end this cycle, yet it has failed so far to persuade anyone it possesses the means to do so.
The Bottom Line
The attempt by OpenAI to build an independent advisory group has fallen apart, bringing confusion with it. The company remains without much gained from past errors.
What follows is the larger concern. OpenAI holds back a stack of findings waiting to see light, and mathematicians are already preparing for a wave of discoveries to hit. Every new release risks another round of arguments over who deserves the credit and contests over who got there first.
Instead of viewing mathematicians as spectators, OpenAI should treat them as partners on equal footing. It requires giving AGMAI room to carry out its work at a steady pace, and it requires being open about what the group’s role actually is. Otherwise, the pattern repeats itself: outstanding work, then disorder, then apologies, then another announcement.
The firm has earned a name for acting quickly, and it must now learn to act with care instead.
Source material: “OpenAI keeps bulldozing mathematicians,” The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.