OpenAI cheated on career‑defining math problem, says NYU mathematician – TechCrunch
openai
| Source: Mastodon | Original article
OpenAI allegedly obtained insider information on a breakthrough math problem and claimed a full proof before the researchers could publish their results.
OpenAI has been accused of “fighting dirty” over a high‑profile mathematics problem, a claim made by NYU mathematician Tristan Buckmaster in a new TechCrunch report. Buckmaster and his collaborator, Alpöge, say they learned that details of their own work on the Navier–Stokes Millennium Prize Problem were passed to OpenAI while they were still finalising results. When they reached out to the company, OpenAI replied that it had already produced a complete proof of the central problem.
The allegation adds a fresh twist to the controversy that erupted earlier this month when OpenAI announced that an internal reasoning model had generated a 165‑page solution to the Navier–Stokes existence and smoothness question, complete with a formal Lean proof. Mathematicians have long questioned the claim because the proof has not been made public and because the approach taken by Buckmaster and Alpöge – a less common line of attack – appears to have been mirrored by the company at the same time.
Why it matters is twofold. First, the Navier–Stokes problem is one of the seven Clay Millennium prizes, carrying a million‑dollar reward and profound implications for fluid dynamics. A genuine solution would be a landmark scientific breakthrough. Second, the dispute raises broader concerns about how large AI firms use proprietary or user‑generated data, and whether they can claim credit for discoveries that may have been sourced from external researchers.
What to watch next: independent verification of OpenAI’s proof, likely through peer‑reviewed publication or open release of the Lean code; further statements from Buckmaster, Alpöge and other mathematicians; and any legal or policy responses concerning data use in AI‑assisted research. As we reported on 8 September 2026, OpenAI’s claim of a “significantly more capable than GPT‑6 Astra” model solving Navier–Stokes sparked intense debate – this latest accusation could intensify scrutiny of the company’s research practices.
Sources
Back to AIPULSEN