Twenty-five Fields Medal holders accuse OpenAI: the proof gets published before it is verified, and the collaborator goes unmentioned

Twenty-five mathematicians holding the Fields Medal - the award mathematics regards as its equivalent of the Nobel in physics - have signed an open letter against the artificial intelligence labs. Not against the machines. Against the way they treat people's work.

The objection is specific and has nothing to do with fear of robots. Solutions, the signatories say, are published in a rush, with no time for a proper write-up, without separating out the new methods and ideas, and without citing other people's prior work. "As in all creative professions, this raises serious questions about attribution of credit and about plagiarism," they write. Who did the work and who signed it - that is the dispute.

This week New York University professor Tristan Buckmaster alleged that OpenAI pressured him not to name as a collaborator a person working for Anthropic who took part in solving an important mathematical problem. Buckmaster went a step further: he asked whether the company had used their work with Codex to produce a proof of its own - over a single marathon weekend of computation. That OpenAI proof has, so far, still not been confirmed.

On Thursday the company withdrew from sponsoring a mathematics event at CalTech, after researchers from the university criticised it. A quiet withdrawal with no explanation - a move that says more than a statement would.

What is worth noticing is the mechanics beneath the whole story. If a lab sees that a usable path leads to a discovery, it can spend tens of millions of dollars on language models and reach the proof before the researcher who has worked on the problem for years. That is not a contest of intelligence but of budget. And the logical consequence is that mathematicians will start staying quiet about what they are working on - the exact opposite of the open research culture they spent decades building.

The letter follows the Leiden Declaration from June, a document by a working group of mathematicians examining the same questions and offering recommendations for researchers, institutions and policymakers. Both documents defend the same thing: the value in mathematics is not only in the proof and in whose name sits under it, but in the structure around it - which educates students, finds new questions and folds them into shared knowledge.

The signatories frame it as a question reaching far beyond their profession: how do we make sure that, while artificial intelligence changes the way work is done, we do not lose sight of why the work existed in the first place. Mathematicians are first only because their work is the easiest to measure - a proof either holds or it does not. Professions where the result is harder to verify will have the same conversation later, and without 25 medallists to start it.