# OpenAI Claims Historic Math Breakthrough, But Researchers Allege Data Theft > OpenAI solved the Navier-Stokes math problem using an unreleased AI model, but researchers accuse the company of training on their private Codex data. - Canonical URL: https://coreiten.com/en/article/openai-claims-historic-math-breakthrough-but-researchers-allege-data-theft - Language: en - Section: Tech News - Author: Sami - Published: 2026-09-09T16:03:04+03:00 - Modified: 2026-09-09T16:03:04+03:00 - Publisher: CoreITen (https://coreiten.com) - Keywords: OpenAI Navier-Stokes solution, GPT-6 Astra, Tristan Buckmaster, AI math breakthrough, Codex data privacy, Millennium Prize Problem ## Summary OpenAI claims its unreleased AI model solved the historic Navier-Stokes problem, but academics immediately alleged data theft involving cloud-based tools. - OpenAI deployed 10,000 concurrent AI agents using a powerful internal model to tackle the 90-year-old mathematical problem. - The Navier-Stokes equation is one of the seven Millennium Prize Problems, carrying a $1 million reward that OpenAI confirmed it will not claim. - New York University professor Tristan Buckmaster and Anthropic researcher Levent Alpöge alleged OpenAI's proof utilized routes developed using Codex and Claude. - OpenAI technical staff member Sébastien Bubeck stated the company did not see the researchers' work until it was publicly released. - OpenAI conceded that de-identified data derived from product usage could have potentially helped improve its models. **Why it matters:** This controversy exposes a major trust vulnerability where academics fear cloud-based AI tools might absorb and repackage their unreleased proprietary drafts. --- OpenAI has claimed a historic milestone by discovering a solution to the Navier-Stokes mathematical problem, but the achievement is already mired in controversy. The 90-year-old problem, which governs the flow of liquids and gases, was reportedly solved using an unreleased internal AI model that is significantly more powerful than the newly launched GPT-6 Astra. To achieve this, the company deployed 10,000 concurrent AI agents. The Navier-Stokes equation is one of the seven Millennium Prize Problems, carrying a $1 million reward for a verified solution. In [a blog post on Tuesday](https://openai.com/index/navier-stokes-solution/), OpenAI announced the breakthrough, noting that training for this specific internal model began on August 28th. The company stated the model exhibited unprecedented performance in mathematics benchmarks, though OpenAI confirmed it does not plan to claim the $1 million prize. ### The Codex Data Dispute The celebration was immediately challenged by the academic community. Just one day prior to OpenAI's announcement, New York University mathematics professor Tristan Buckmaster and Anthropic researcher Levent Alpöge published their own findings on a related problem. Buckmaster claims that OpenAI's proof utilizes a specific route that he and Alpöge had been actively developing using OpenAI's Codex and Anthropic's Claude. > I asked whether the model had been trained on, or had access to, our sessions in Codex, into which we had been putting all our drafts for the whole of this project. I was told the model did not look up user data. I asked again, about training, and I did not get an answer. > > - Tristan Buckmaster, New York University OpenAI is actively attempting to defuse the allegations of data scraping. The company stated that no specific user data was accessed to solve the problem, though it conceded that de-identified data derived from product usage could have helped improve its models. OpenAI pointed to [its statement on X](https://x.com/OpenAI/status/2097375276384567642?s=20) to reiterate its defense. Sébastien Bubeck, a member of the technical staff at OpenAI, [similarly said](https://x.com/SebastienBubeck/status/2097379411691516310?s=20) that the company did not see any of the researchers' work until it was released publicly. "One can in hindsight see that our proofs differ significantly and even the precise results proved are different," Bubeck explained. Buckmaster quickly fired back on Mastodon, arguing that OpenAI is openly admitting to using training data from a period after the researchers had already found their result. ### The Blurry Line Between Training and Theft This dispute highlights a critical vulnerability in the modern scientific process. If researchers cannot trust cloud-based AI tools like Codex with their unreleased drafts, the academic community may be forced to abandon these platforms entirely. The timeline here is incredibly tight, and OpenAI's admission that "de-identified data" might have contributed to the model's baseline intelligence does little to reassure academics who feel their proprietary logic was absorbed and repackaged. OpenAI's decision to walk away from the $1 million Millennium Prize is likely a calculated strategic move rather than pure altruism. Claiming the prize would invite intense, formal scrutiny from the Clay Mathematics Institute regarding the exact provenance of the proof. By declining the money, OpenAI secures the PR victory of the mathematical breakthrough while sidestepping a grueling intellectual property investigation that could expose exactly how its internal models ingest user data. ## Sources - [theverge.com](https://www.theverge.com/ai-artificial-intelligence/991710/openai-navier-stokes-solution)