An OpenAI researcher allegedly pressured mathematician Tristan Buckmaster to remove an Anthropic co-author from a paper documenting progress on the Navier-Stokes equations, one of mathematics' most famous unsolved problems. When Buckmaster refused, the researcher threatened him, according to Buckmaster's account. OpenAI subsequently announced its own breakthrough using the same solution approach that Buckmaster had been developing.
Buckmaster, a mathematician at the University of Maryland, claims he shared his work-in-progress drafts through OpenAI's Codex tool while working on the Navier-Stokes problem. The equations govern fluid dynamics and remain one of the Clay Mathematics Institute's seven Millennium Prize Problems worth $1 million each. After word of Buckmaster's progress reached OpenAI internally, a company researcher contacted him requesting he remove his Anthropic-affiliated co-author from the paper, citing competitive concerns.
When Buckmaster declined, the situation escalated. The researcher allegedly threatened Buckmaster for refusing to drop the co-author. Despite these pressures, Buckmaster maintained the original paper with all listed collaborators intact. Separately, OpenAI then published its own mathematical breakthrough that traced the same solution path Buckmaster had been pursuing.
The core issue centers on data access and competitive practices within AI companies. Buckmaster had uploaded intermediate drafts to Codex, trusting OpenAI's assurances that the model would not access user data for company benefit. However, the timing and specifics of OpenAI's subsequent breakthrough raise questions about whether information from Buckmaster's research influenced the company's work.
This episode illuminates broader tensions between AI research transparency and competitive incentives. Researchers often share drafts publicly or through collaborative tools to accelerate progress. Yet when researchers work with tools owned by competing companies, information asymmetries emerge. OpenAI's Codex, a code and text generation model, processes vast amounts of user input. The dispute hinges on whether data from Buckmaster's uploads influenced OpenAI researchers who subsequently worked on the same problem.
The incident also exposes internal dynamics at OpenAI that prioritize competitive advantage over collaborative norms. Attempting to coerce a researcher into removing a co-author from a paper crosses ethical boundaries common in academic publishing. Threatening behavior when that coercion fails further compounds the concern.
Anthropic, founded by former OpenAI members, competes directly with OpenAI on large language models and AI safety research. This rivalry adds context to the pressure campaign. An Anthropic-affiliated researcher on a major mathematical breakthrough would elevate that company's profile in AI-driven discovery.
OpenAI has faced previous scrutiny regarding researcher independence and competitive conduct. This latest account from Buckmaster, a respected mathematician, adds specificity to those concerns. The company's ability to influence users through direct contact, combined with access to their data, creates structural advantages that merit examination.
The Navier-Stokes problem represents a frontier where AI-assisted mathematics could generate genuine breakthroughs. Buckmaster's work demonstrates legitimate progress. OpenAI's parallel achievement using the same method suggests either independent discovery or influence from information Buckmaster had shared. Establishing clear ethical norms around researcher conduct and data handling becomes essential as AI companies increasingly compete on research output.