OpenAI forms a math advisory group as its models crack 100 open problems

The lab is standing up a largely unpaid, independent advisory body of mathematicians — a bid to manage the optics and the fallout of AI now resolving long-standing open problems, while explicitly refusing them any say over the pace of its research.

Published: 2026-09-22 Category: Quick Take

What happened

OpenAI has created a mathematics advisory group whose stated purpose, per its own blog post, is to "serve as a bridge to the mathematical community and broader public, giving mathematicians a voice in how we move forward." The announcement lands alongside the claim that OpenAI's models have now resolved more than 100 open problems in mathematics — a threshold that shifts the conversation from "can AI do math" to "what does it mean for the discipline when AI does math, and who decides what gets published."

The structure is careful, and telling

The group is framed as an advisory layer rather than a gatekeeper. Members will assess the significance of new results and help coordinate their release, but they won't be paid, they can offer unsolicited advice, go public with their own views, and control their own membership. That buy-in is designed to give the body a measure of independence from the company.

Yet the limits are drawn just as deliberately. As the blog post puts it, "the group will not be responsible for advising us on how to pace our internal progress on mathematics." In other words: mathematicians get a voice on what AI has found, but not on whether OpenAI should slow down looking for more. That carve-out matters — it's the same "you can audit, you can't steer" tension that keeps recurring across frontier labs on everything from safety to copyright.

Why the composition matters

Nine prominent mathematicians have been named as initial members. Notably, only one of them — IAS's Camillo De Lellis — also signed the earlier Fields Medalists' letter, the open letter that raised concerns about AI's rapid encroachment into mathematical research. That is a deliberate signal. OpenAI is reaching past its most vocal critics while still bringing one of them inside the tent, hedging between credibility and control.

The sharp read

This is OpenAI managing a legitimacy problem, not just convening experts. When an AI resolves a hundred open problems, the scarce resource stops being the results and becomes the interpretation — who says a proof is real, who gets credit, who decides a result is worth the community's attention. By funding a body that looks independent but cannot slow the underlying research engine, OpenAI gets the credibility of external oversight and the freedom of none. It is a bridge, as advertised, but bridges carry traffic in both directions, and this one is engineered to let influence flow to the community while keeping control of the throttle firmly in-house.

Source: TechCrunch