Researchers Challenge OpenAI's Navier-Stokes Proof and Transparency
OpenAI released hundreds of mathematical proofs on GitHub, but researchers argue its solution to the Navier-Stokes problem lacks transparency and formal accuracy.
Researchers and academic institutions are challenging the validity of a reported solution to the Navier-Stokes problem by OpenAI. On October 6, 2026, the company released over 370 mathematical proofs on GitHub covering algebra, theoretical computer science, and mathematical logic. This release followed OpenAI's claim of solving the Navier-Stokes blow-up, a $1 million Millennium Prize challenge.
In a preprint published October 6, researchers Alexander Bastounis, Fabian Circelli, and Anders C. Hansen argued that the Lean formalization OpenAI provided does not correspond to its natural-language mathematical argument. They contend that while a proof assistant can certify a formal argument is mechanically valid, it cannot guarantee the autoformalization process preserved the original meaning of the prose. Researchers from the University of Cambridge and King's College London identified similar discrepancies, raising concerns that AI-generated proofs may lack genuine originality or human understanding.
The Advisory Group on Mathematics and Artificial Intelligence (AGMAI), hosted by the Institute for Advanced Study, criticized OpenAI for failing to meet transparency standards. The group and other experts argue that the use of proprietary models and the withholding of prompt histories prevent the broader mathematical community from properly verifying the results.
