OpenAI announced that it had solved the Navier-Stokes problem, which had remained unsolved for approximately 90 years. The problem concerning the flow of liquids and gases is one of the seven Millennium Prize Problems, and there is a $1 million prize for its solution. In its announcement dated September 8, 2026, the company said it had used 10,000 simultaneous agents with an internal artificial intelligence model more powerful than GPT-6 Astra. It was stated that training of the model had begun on August 28 and that it had demonstrated unprecedented performance in mathematics. OpenAI does not plan to claim the prize.
However, the announcement sparked controversy. New York University mathematics professor Tristan Buckmaster and Anthropic researcher Levent Alpöge published their findings on a related problem one day before OpenAI’s announcement. Buckmaster said OpenAI had produced a proof based on an approach developed by his own teams using Codex and Claude, and that he had contacted the company about their progress. He also said he had asked whether OpenAI had accessed their drafts in Codex sessions and was told that user data had not been examined, but that he had received no answer about the use of the data for training.
OpenAI stated that no specific user data had been accessed for the solution, but that it could not be ruled out that de-identified data obtained from product use may have contributed to the development of the models. Sébastien Bubeck argued that the teams had not seen the work before it was published and that the proofs were different. Buckmaster, meanwhile, claimed that OpenAI had acknowledged using training data obtained after their results.
Why it matters
This development raises the question of whether AI systems can be regarded not merely as computational tools, but as actors that produce solutions in mathematical research. However, acceptance of such a claim depends on the evidence presented being examined by independent mathematicians and on determining whether it meets the criteria of the problem. Another dimension of the debate concerns how academic contribution should be defined in terms of the studies and tools through which the result was developed. Although OpenAI stated that specific user data had not been accessed, its refusal to rule out the possibility that de-identified usage data contributed to model development leaves open the question of the evidence’s source and the use of data. For this reason, the issue concerns academic circles not only in terms of the mathematical result, but also with regard to verification, transparency, and recognition of researchers’ contributions.
Background
OpenAI is not a new name in the FikirPilot archive: we have published 30 news articles mentioning the name in the last 90 days; the most recent is dated September 11, 2026.