Twenty-five mathematicians, including several Fields Medal winners, signed an open letter warning that AI labs racing to solve famous problems are undermining attribution and documentation norms, according to the letter and TechCrunch.
The letter says AI-announced solutions often skip "proper writeup, the isolation of new methods and ideas, and citing relevant previous work," raising what the signatories called "severe attribution and plagiarism questions." They warned that without mathematicians involved in developing and integrating AI-generated ideas, the field's normal process for building on prior work would break down.
The letter follows an accusation by NYU mathematician Tristan Buckmaster, who said Sept. 8 that OpenAI pressured him to withhold credit from an Anthropic-affiliated collaborator after OpenAI announced its agents had solved the Navier-Stokes equations using 10,000 agents over 88 hours, according to MIT Technology Review. Buckmaster called it "a Deep Blue-Kasparov moment" for mathematics. OpenAI withdrew its sponsorship of a Caltech math event Sept. 11 after criticism from researchers there, TechCrunch reported.
The mathematicians are not disputing that AI can solve hard problems. Their complaint is about credit, and about whether announcements are outrunning the peer review needed to catch errors and assign it fairly. Builders selling AI-discovered results as scientific output should expect the same scrutiny mathematicians are now demanding, on timelines that don't match a product launch calendar.