On September 11, twenty-five mathematicians who have won the Fields Medal — the top prize in math — signed an open letter criticizing how AI labs handle major mathematical discoveries. The letter follows a week where NYU professor Tristan Buckmaster accused OpenAI of pressuring him not to credit a colleague at Anthropic for solving an important problem, then racing to publish its own proof over a single weekend.
The letter argues that AI-generated proofs are being announced before anyone has time to verify them, write them up properly, or credit the human mathematicians whose prior work made them possible. OpenAI's recent proof remains unverified. The company also withdrew its sponsorship of a CalTech math event after facing criticism from researchers.
Why it matters beyond math: When AI can spend millions of dollars on inference to beat a human researcher to a result, the incentive structure breaks. Researchers who share their work openly risk getting scooped by a model that was trained on their ideas. The letter warns that this will push scientists toward secrecy.
The mathematicians are not saying AI should stop solving problems. They are saying the solutions need to be understandable, verifiable, and properly attributed — or the entire system of human knowledge transmission collapses. Your field is next.