25 Mathematicians vs OpenAI: The Fight Over Who Owns Mathematical Discovery

Twenty-five leading mathematicians have signed an open letter arguing that AI labs are threatening their intellectual work, and the specific target is OpenAI. This isn't the first time the two sides have clashed, but the escalation signals that a simmering dispute about the nature of mathematical creativity and AI's role in it has reached a new level of public confrontation.

The feud has roots in OpenAI's repeated claims that its models have made significant progress on hard mathematical problems, including the claim, reported here earlier this month, that GPT-6 had solved a 90-year-old maths problem. That claim was contested almost immediately by mathematicians who argued that what the model had actually done was either already known, subtly wrong, or not a genuine proof in any meaningful sense.

What the Mathematicians Are Actually Arguing

The open letter, according to TechCrunch, centres on the argument that AI labs are claiming mathematical achievements in ways that misrepresent how mathematical discovery actually works, and that those claims are damaging both to public understanding and to the careers and recognition of working mathematicians.

Mathematics has a very specific culture around proof and credit. A result is not a result until it has been verified by the mathematical community through peer review. Claims of discovery that haven't gone through that process are, in the mathematical tradition, not claims of discovery at all. They're conjectures at best.

AI labs, including OpenAI, have a habit of announcing mathematical achievements through press releases and blog posts that bypass this process entirely. When a model produces something that looks like a proof of a significant result, the temptation to announce it immediately is commercially understandable. The mathematical community, however, finds this practice somewhere between irritating and actively harmful.

The specific concern in the letter, as reported, is that AI labs are threatening mathematicians' intellectual work. This phrasing covers at least two distinct anxieties. The first is credit: if an AI produces a result that a mathematician was working toward, who gets recognition? The second is methodology: if AI-generated proofs that haven't been properly verified start circulating, they pollute the mathematical literature in ways that are genuinely difficult to unpick.

OpenAI's Position

OpenAI has not been shy about its mathematical ambitions. The company has framed progress on hard mathematical problems as one of the clearest demonstrations of frontier model capability, and in competitive terms, mathematical benchmarks are clean and legible in a way that more subjective capability assessments are not. If your model can solve IMO problems or make progress on the Riemann Hypothesis, that's a headline.

But the company has also, arguably, been sloppy about the distinction between a model producing output that resembles a proof and a model actually proving something. These are not the same thing, and the difference matters an enormous amount to anyone who works in the field.

Why This Is More Than Academic Turf War

It would be easy to dismiss this as mathematicians being protective of their territory. But there are real stakes here that extend beyond professional pride.

Mathematical proofs underpin enormous amounts of cryptography, computer science, and physics. A claimed result that turns out to be wrong, if it propagates through the literature before being caught, can waste years of other researchers' time and occasionally lead to real-world systems being built on faulty foundations. The verification culture in mathematics is not pedantry. It's quality control for one of the most consequential knowledge systems humans have built.

The question of credit also matters more than it might seem. Mathematics is a field where a career can be built on a single significant result. If AI systems start claiming credit for discoveries that mathematicians have been building toward over years of work, the incentive structures of the entire field shift. Funding bodies, universities, and journals will need to develop frameworks for attribution that simply don't exist yet.

Twenty-five signatories is a meaningful number for an open letter in mathematics. This dispute is not going away, and the next time OpenAI announces a major mathematical result without going through the community's verification process, the response is likely to be even sharper.