Twenty-five Fields Medalists warn frontier AI labs threaten research integrity
An open letter signed by twenty-five Fields Medalists accuses artificial intelligence laboratories of plagiarism, bypassing peer review, and threatening the human transmission of mathematical knowledge.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- Twenty-five leading mathematicians, including multiple Fields Medalists, have signed an open letter warning that frontier AI labs are compromising the integrity of academic research.
- The scholars accuse labs of using 'marathon inference' to claim mathematical breakthroughs without undergoing traditional peer-review processes.
- The conflict intensified after NYU professor Tristan Buckmaster accused OpenAI of pressuring him to omit credit for Anthropic researcher Levent Alpöge on a Navier-Stokes proof.
A coalition of twenty-five Fields Medalists—the highest honour in the mathematical sciences—has published a scathing open letter accusing leading artificial intelligence laboratories of threatening the structural integrity, attribution standards, and human transmission of mathematical research. The letter reflects a massive escalation in the conflict between elite academia and Silicon Valley, following a series of controversial AI-generated proofs claimed by frontier laboratories. The signatories warn that the tech industry's rush to deploy 'marathon inference' loops to solve historic problems is actively undermining peer-review systems and introducing toxic plagiarism and priority disputes into the discipline.
The immediate catalyst for the open letter was a public priority dispute involving New York University mathematics professor Tristan Buckmaster. Buckmaster accused OpenAI of aggressively pressuring him to omit credit for Levent Alpöge, an AI researcher and mathematician at Anthropic, who collaborated on a key foundation of OpenAI’s computer-verified Navier-Stokes blowup proof. The letter alleges that OpenAI researchers became aware of Buckmaster and Alpöge's impending Euler equation breakthrough and accelerated their own 10,000-agent model runs, subsequently claiming academic priority while failing to properly attribute the conceptual groundwork laid by the human researchers.
Beyond individual disputes, the mathematicians argue that AI-generated proofs, which often span millions of lines of machine-generated interactive theorem code like Lean, are impossible for humans to peer-review in any traditional sense. By substituting deep conceptual understanding with massive, brute-force computational searches, laboratories are bypassing the intellectual transmission chain that has sustained mathematics for millennia. In response to mounting criticism from academic staff, OpenAI has quietly withdrawn its financial sponsorship of a major upcoming mathematics conference at the California Institute of Technology (Caltech), signaling a widening gulf between frontier labs and the scientific community.
This dispute shows that AI is not just a tool for scientific discovery but an active disruptor of academic governance. If private AI laboratories can bypass peer review and claim priority on historic problems like the Navier-Stokes existence and smoothness problem through brute-force compute, they risk decoupling mathematical truth from human understanding. For research institutions and funding bodies, this letter will likely mandate strict new licensing and attribution guidelines for machine-assisted proofs, ensuring that the human-centered process of mathematical transmission is not erased by opaque corporate compute runs.