25 Fields Medal Winners Have Signed a Letter Challenging OpenAI's Maths Claim

Key points

• 25 Fields Medal recipients, the highest honour in mathematics, signed an open letter warning of a "severe misalignment" between AI companies and the mathematics community.
• OpenAI announced in early September 2026 that its AI had solved a major mathematics problem that had gone unsolved for nearly 100 years.
• The letter's signatories include some of the most decorated living mathematicians, making this one of the most credible pushbacks any AI lab has received from a scientific field.
• AI2Day has been tracking this dispute since 9 September 2026; our reporting found allegations of academic scooping and data misuse well before the open letter landed.

OpenAI made a striking claim in early September 2026: its AI had cracked a mathematics problem that human researchers had failed to solve for close to a century. Twenty-five recipients of the Fields Medal, the prize widely considered the Nobel of mathematics, co-signed an open letter calling this a "severe misalignment" between AI companies and the mathematics community. That's not a polite grumble. It's the field's most decorated minds saying, collectively, that something has gone wrong.

Colva Roney-Dougal, a professor of pure mathematics at the University of St Andrews, spoke to The Guardian about the implications. The concern isn't just about one result. It's about whether AI companies understand what mathematical proof actually means, and whether they're describing their results accurately to a public that has no way to check. One phrase circulating among critics: "immature playground boasting."

What did OpenAI actually claim?

OpenAI said its AI solved a significant open problem in mathematics, one that had resisted human effort for roughly a century. The company hasn't specified which problem, which makes independent verification harder than it should be.

This matters financially as much as scientifically. OpenAI charges businesses and consumers for access to its AI tools, and each headline like this strengthens the case that those tools are worth paying for. Claims that can't be cleanly checked by outsiders are useful for marketing in ways that verified, peer-reviewed results are not.

The transparency problem runs deeper than one announcement. A second mathematician accused OpenAI of dishonesty over its AI maths breakthroughs in our 10 September 2026 story: Andreas Thom said the company never clarified whether his ChatGPT conversations had shaped the very result that built on his own research, and that its answers were "materially misleading." That's a credibility record worth keeping in mind.

Should you be worried about what this means for human researchers?

For working mathematicians, the honest answer is: it depends what the AI actually did. If it found a genuine, verifiable proof, the field will absorb it and move to harder problems, as it always has. Overstated framing is a credibility problem for AI labs, not a defeat for human mathematicians.

For everyone else, the letter is a useful reminder. AI companies have strong financial reasons to describe results in the most impressive terms possible. Scientists whose life's work sits in these fields have equally strong reasons to read the fine print. When those two groups publicly disagree, the right instinct is to notice who is saying what, and who is paying whom.

We've published 12 stories tagged "mathematics" since July 2026, and the pattern holds: results that stand up tend to be described carefully; results announced in sweeping terms tend to need revision.

The practical move is narrow and doable. When an AI company announces a scientific breakthrough, check whether a peer-reviewed paper exists and whether independent researchers in that field are saying the same thing. If the only voices calling it historic are inside the company, wait a week.