The Math World's AI Intellectual Property War

NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge recently published findings on the fluid dynamics of air and water, work that utilized OpenAI and Anthropic models. Following their discovery, Buckmaster alleged that OpenAI staff, specifically Sebastian Bubeck, claimed the company had already solved the $1 million Navier-Stokes problem internally. Buckmaster further suggested that OpenAI attempted to influence the publication process and requested the removal of Alpöge, who is employed by a competitor.

OpenAI denied these allegations, with Bubeck characterizing the claims as inflammatory and asserting that the company's internal research was motivated by external rumors rather than the duo's inputs. While OpenAI officially announced a solution to the complex math problem on Tuesday, it acknowledged that it could not definitively rule out that de-identified data from user interactions—including those of Buckmaster—may have inadvertently improved their models. This tension highlights a critical reality: under standard terms of service, AI companies retain the right to train on user-submitted content unless an explicit opt-out is exercised. As AI becomes an essential tool for scientific discovery, the lack of transparency regarding how these models integrate private research threatens to undermine the foundations of academic credit.