OpenAI announced that it had solved the Navier-Stokes problem in 88 hours with a system in which approximately 10 thousand artificial intelligence agents worked together, using a model it has not yet made available. The problem is one of the seven Millennium Prize Problems and is awarded 1 million dollars by the Clay Mathematics Institute. The announcement sparked debate over when the company turned to the problem and about academic transparency. One day before the announcement, New York University professor Tristan Buckmaster and Levent Alpöge, who works at Anthropic, had published related studies; Alpöge was not conducting the research on behalf of Anthropic. Buckmaster said that OpenAI was aware of the studies and asked about their start date and training data, and that in a meeting, his decision to risk his career was questioned. OpenAI argued that specific user data had not been used; it did not rule out the indirect contribution of de-identified data. Buckmaster also raised demands for Codex access, attribution, and Alpöge’s removal from the list of co-authors. Sébastien Bubeck disputed some of the claims. No definitive assessment can be made because the parties’ accounts contradict each other.
Why it matters
This development shows that in AI-assisted mathematical research, not only the result but also the independent verification of the solution, the traceability of contributions, and how academic work is allocated are decisive. Since a model that has not yet been made available and a multi-agent system are involved, researchers’ ability to reproduce the method and scrutinize the claim currently appears limited. Disputes over training data, the possible indirect contribution of de-identified user data, attribution, and co-authorship directly concern the balance of access and power between university researchers and AI companies. The key open questions are whether the work will be considered valid under the Millennium Prize and how the impact of previous studies on the result will be documented.
Background
OpenAI is not a new name in the FikirPilot archive: we have published 28 news reports mentioning it in the last 90 days; the latest is dated September 9, 2026.