OpenAI Forms Math Advisory Group After Its AI Solves 100+ Unsolved Problems

OpenAI has formed a new advisory group on mathematics and artificial intelligence after an internal AI model reportedly resolved more than 100 previously unsolved mathematical problems, including a claimed solution to the Navier-Stokes equations, one of the seven Millennium Prize Problems that have stumped mathematicians for decades.

The announcement, made September 21, comes amid growing backlash from the mathematical community, including an open letter signed by 25 Fields Medal winners warning that AI labs racing to solve famous open problems threaten the integrity and culture of mathematical research itself.

What OpenAI is claiming

According to OpenAI, an internal model — not yet publicly named in available reporting — has resolved more than 100 open mathematical problems spanning most major areas of mathematics. The most attention-grabbing of these claims is a solution to the Navier-Stokes existence and smoothness problem, one of the Clay Mathematics Institute’s seven Millennium Prize Problems, each carrying a $1 million reward and decades of unsuccessful attempts by human mathematicians. That solution was reportedly published on September 8, roughly two weeks before OpenAI’s broader announcement of the additional 100-plus solved problems and the new advisory group.

Why an advisory group, and why now

OpenAI’s new Advisory Group on Mathematics and Artificial Intelligence is hosted at Princeton’s Institute for Advanced Study, one of the most prestigious research institutions in the world for pure mathematics and theoretical physics. In its own words, OpenAI said the group “will serve as a bridge to the mathematical community and broader public, giving mathematicians a voice in how we move forward.” The timing — arriving alongside the announcement of the 100-plus solved problems rather than before it — has drawn its own scrutiny, with critics suggesting the advisory group looks at least partly like a response to mounting criticism rather than a structure built proactively ahead of the controversy.

The Fields Medal backlash

The most pointed criticism of OpenAI’s approach has come from an open letter signed by 25 Fields Medal winners — recipients of mathematics’ most prestigious honor — who argue that AI labs competing to solve famous, long-standing problems threaten meaningful intellectual work in the field. The letter reflects a deeper anxiety within the mathematical community: that a research culture built around slow, deliberate human problem-solving, often over the course of entire careers, could be disrupted or devalued if AI models can solve in weeks or months what has taken humans generations to even approach.

A complicated relationship with the advisory group itself

Notably, only one member of OpenAI’s new advisory group, mathematician Camillo De Lellis, is also among the 25 Fields Medal winners who signed the protest letter. That overlap — or rather, the lack of it — has raised questions about how representative the advisory group actually is of the concerns driving the backlash, and whether mathematicians most skeptical of OpenAI’s approach were meaningfully included in shaping the group’s composition and mandate.

What the advisory group can and can’t do

Perhaps the most significant detail in OpenAI’s announcement is what the advisory group is explicitly not empowered to do. The Institute for Advanced Study, which hosts the group, stated plainly: “Although we will give advice, we do not have decision making power at any AI company.” In other words, the group can offer perspective, raise concerns and provide a channel for mathematician input, but it has no formal authority to slow down, redirect or otherwise constrain OpenAI’s research agenda around mathematics. Critics have seized on that limitation as evidence that the advisory group functions more as a public relations response than a genuine check on OpenAI’s approach.

Why the Navier-Stokes claim matters so much

The Navier-Stokes equations describe the motion of fluid substances, and the question of whether smooth, well-behaved solutions always exist for these equations in three dimensions has remained unresolved since the Clay Mathematics Institute listed it as one of seven Millennium Prize Problems in 2000. Only one of the seven, the Poincaré conjecture, has been solved and formally verified in the years since. A legitimate AI-generated solution to Navier-Stokes would represent one of the most significant developments in the history of mathematical problem-solving — but it would also require an extraordinary level of independent verification before the broader mathematical community accepts it as valid, a process that has not yet concluded.

The verification problem

OpenAI has not released the specific list of the 100-plus problems its model claims to have solved, a gap that has fueled skepticism among mathematicians who note that claims of this magnitude require rigorous, transparent peer review before they can be taken at face value. Without a public accounting of exactly which problems were solved and how, independent mathematicians have limited ability to verify OpenAI’s claims directly, leaving much of the current discussion dependent on OpenAI’s own characterization of its model’s achievements rather than confirmed, peer-reviewed results.

What this means for the AI industry more broadly

OpenAI’s announcement lands amid intensifying competition among major AI labs to demonstrate mathematical reasoning capability, an area widely seen as a proxy for broader advances in AI reasoning that could translate into other high-value domains like scientific research, engineering and software development. That competitive pressure is part of what has fueled concern among mathematicians that labs are racing toward headline-grabbing claims about famous problems rather than pursuing more incremental, verifiable contributions to the field, a dynamic the Fields Medal winners’ letter explicitly calls out as corrosive to how mathematical research has traditionally been conducted and evaluated.

How the mathematical community is likely to respond

Given the scale of skepticism already expressed by prominent mathematicians, OpenAI’s claims are likely to face sustained scrutiny in the coming months as the broader research community attempts to independently verify or challenge the Navier-Stokes solution and the other reported results. How the newly formed advisory group navigates that scrutiny — and whether it can build credibility with mathematicians who remain skeptical of both the claims and the group’s limited authority — is likely to shape how much trust the mathematical community places in OpenAI’s future announcements in this space.

What happens next

OpenAI has not indicated a timeline for releasing the full list of problems its model claims to have solved, nor has it detailed what role, if any, the new advisory group will play in that disclosure process. In the meantime, the Navier-Stokes claim in particular is likely to draw intense attention from the mathematics community, given both the size of the associated prize and the problem’s decades-long status as one of the most difficult open questions in the field. Whether OpenAI’s claims hold up to that scrutiny will be a major test of how much weight the mathematical community is willing to give AI-generated proofs going forward.

How AI models approach mathematical proofs

Modern AI reasoning models tackle mathematical problems very differently from how they generate ordinary text, typically relying on extended chains of step-by-step logical reasoning, sometimes combined with formal proof-checking systems that can verify whether each individual step in a proof follows validly from the last. That combination — a powerful language model generating candidate reasoning paths, paired with more rigid formal verification tools — is part of why AI labs have made rapid progress on certain categories of mathematical problems in the past two years, even as fully autonomous solutions to problems as difficult as Navier-Stokes remain, by most mathematicians’ accounts, at the very outer edge of what current systems should plausibly be able to achieve without significant human guidance.

Skepticism from outside the advisory group

Mathematicians who have publicly commented on OpenAI’s claims, including several who did not sign the Fields Medal letter, have expressed a mix of cautious interest and continued skepticism, noting that AI-assisted proofs have in the past occasionally contained subtle errors that only became apparent after extended peer review, sometimes months or years after an initial claim generated headlines. That history has left many in the field inclined to treat OpenAI’s unreleased list of 100-plus solved problems as a provisional claim rather than a settled result, at least until independent mathematicians outside OpenAI have had the opportunity to examine the specific proofs in detail.

The Millennium Prize context

The Clay Mathematics Institute’s seven Millennium Prize Problems, established in 2000, were each considered so difficult that solving any one of them would represent a landmark achievement in mathematics, with a $1 million reward attached to each as an incentive. Only the Poincaré conjecture, solved by Grigori Perelman in the early 2000s, has been fully resolved and formally accepted by the mathematical community among the original seven. If OpenAI’s Navier-Stokes claim withstands scrutiny, it would mark just the second of the seven problems to be solved in the roughly quarter-century since the prizes were established — a milestone significant enough to explain both the intensity of interest in the claim and the equally intense demand for rigorous verification before it’s accepted.

What Princeton’s involvement signals

OpenAI’s choice to house its new advisory group at Princeton’s Institute for Advanced Study, rather than within OpenAI’s own organizational structure, appears designed to lend the group a degree of independence and academic credibility that an internal committee might lack. The Institute has a long history as a home for foundational mathematical and theoretical research, and its willingness to host the group suggests at least some segment of the elite mathematics establishment sees value in engaging with AI labs directly rather than remaining entirely outside the conversation, even as other prominent mathematicians, through the Fields Medal letter, have chosen a more confrontational posture toward the same underlying trend.

The broader stakes for AI-assisted research

Beyond mathematics specifically, the controversy around OpenAI’s claims touches on a question that extends across scientific research more broadly: as AI models become more capable of generating novel results in technical fields, how should the scientific and academic communities adapt their verification, credit and peer-review processes to keep pace, without either rejecting legitimate breakthroughs out of institutional caution or accepting unverified claims simply because they come from a well-resourced AI lab. Mathematics, with its unusually rigorous and formal standards for what counts as a valid proof, has become an early and closely watched test case for how that broader tension plays out.

Follow Smashology Media’s Tech coverage as mathematicians respond to OpenAI’s claims.

Sources

• TechCrunch: OpenAI forms math advisory group as its AI resolves more than 100 open problems
• Ground News: OpenAI forms independent advisory group for AI and mathematics after backlash from hundreds of mathematicians