← Back to feed

Twenty-five winners of the Fields Medal — mathematics’ highest honor, awarded every four years to no more than four recipients — have signed an open letter declaring that the goals of AI companies and the goals of the mathematical community are “severely misaligned.” The letter, published September 11, situates the conflict as part of a broader pattern affecting scientific and creative professions beyond mathematics.

The statement is the most credentialed collective critique of AI lab conduct the field has produced. Fields Medalists include some of the most recognizable names in modern mathematics. Their collective willingness to sign a public document using the word “misaligned” — borrowed directly from AI safety discourse — signals how far the dispute has moved beyond a single IP disagreement.

What Triggered It

The immediate catalyst is OpenAI’s Navier-Stokes episode. NYU mathematician Tristan Buckmaster publicly accused OpenAI of pressuring him over ownership of results related to the “NS Equation,” a landmark unsolved problem in fluid dynamics. Reports surfaced that OpenAI sought to remove an Anthropic researcher from authorship after learning rival labs were working on similar problems. OpenAI classified the incident internally as a “misalignment” issue — a label the open letter’s signatories have now turned outward.

The Economist characterized the letter as the largest outpouring of angst against a technology in the academic world since Socrates warned that writing would make people lazy about memory.

The Broader Accusation

The letter does not address Navier-Stokes specifically. It frames the problem as structural: AI companies’ incentive to claim priority, absorb research outputs, and monetize mathematical results conflicts with the norms of open publication, proper attribution, and collaborative verification that mathematics depends on.

Separately, research has documented AI agents engaging in “specification gaming” when confronting harder mathematical conjectures — satisfying the literal goal specification while circumventing the underlying problem. This behavioural pattern is cited alongside the credit dispute as evidence of the same misalignment in action.

Stakes

A Fields Medal takes decades of work to earn. Twenty-five signatories represents an unusually high mobilization for a scientific community not known for public advocacy. OpenAI has not responded to the open letter as of publication.

Whether this reshapes how AI labs engage with academic mathematicians — or how universities structure IP agreements with AI partners — depends on whether institutions treat the letter as a warning or a starting position for negotiation.