The academic world is reeling from a public clash between OpenAI and mathematicians Tristan Buckmaster and Levent Alpöge. At the center is the Navier-Stokes existence and smoothness problem, one of the Clay Millennium Problems carrying a $1 million prize. Buckmaster accuses OpenAI of pressuring him to remove Alpöge, an Anthropic employee, from their joint paper and potentially using their draft work to train its models. While OpenAI CEO Sam Altman defends his team’s integrity, calling the plagiarism claims unfounded, the company’s own admission creates significant ambiguity.
OpenAI researcher Sébastien Bubeck and CEO Altman state they mobilized resources after hearing rumors that Anthropic’s models had solved the problem. They deny needing hints from the mathematicians, with Boaz Barak noting the model proved a stronger claim independently. Yet, OpenAI’s official blog post concedes a critical uncertainty: “While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models.” This admission undermines the defense that the AI worked in a vacuum, especially since Buckmaster claims he entered similar approaches into the system.
Alpöge disputes Altman’s characterization that he refused to coordinate, stating he was open to collaboration and that authorship was secondary to scientific progress. The disconnect highlights a growing friction between traditional academic norms and the aggressive, resource-heavy tactics of tech giants. Altman insisted the team acted with generosity, but the mathematicians view the interaction as coercive. The incident exposes how quickly private companies can pivot to dominate specific research verticals based on mere internet chatter.
Field medalist Terence Tao warns this dynamic threatens the foundation of open science. If rumors alone trigger massive AI-driven efforts to solve problems before original researchers can publish, scholars may stop sharing early-stage ideas. The incentive structure now punishes transparency. Researchers feeding data into these platforms risk having their own findings used against them, with opt-out settings offering little reassurance. We are witnessing a potential reversal of centuries of collaborative tradition, replaced by a race where silence becomes the only safe strategy for independent thinkers.
