OpenAI stated Tuesday that its internal AI model produced a Lean-verified proof for the Navier-Stokes fluid equations' potential to "blow up," utilizing approximately 10,000 coordinating agents over 88 hours. This achievement addresses one of seven Millennium Prize Problems, each carrying a $1 million award from the Clay Mathematics Institute. The announcement followed a claim by OpenAI that the result came from a next-generation model significantly more capable than GPT-6 Astra.
However, NYU mathematician Tristan Buckmaster released a statement accusing OpenAI researcher Sébastien Bubeck of learning about his and Anthropic researcher Levent Alpöge’s unpublished, related proof before pressuring them regarding credit. Buckmaster alleged that Bubeck offered options including publishing after their team or excluding Alpöge due to his employment at a rival lab. OpenAI, Bubeck, and CEO Sam Altman denied these accusations, with Bubeck sharing text messages he claimed demonstrated good faith and an offer for Buckmaster’s team to publish first.
This incident highlights the growing friction between proprietary AI development and academic research norms as large language models tackle high-stakes mathematical problems. The core tension lies in the opacity of training data usage and the competitive dynamics between major AI labs like OpenAI and Anthropic. While OpenAI asserts independent discovery through massive parallel agent coordination, the proximity of timelines and shared subject matter raises questions about intellectual property boundaries in AI-assisted research.
From a Credibility perspective, the lack of independent verification for OpenAI’s 100-page proof and the unresolved nature of Buckmaster and Alpöge’s most advanced results underscore the need for transparent peer review mechanisms in AI-generated mathematics. Fields Medalist Terence Tao’s acknowledgment of the underlying math as a "remarkable achievement" lends weight to the human researchers' contributions, suggesting that current AI capabilities may still rely heavily on foundational human insight. Stakeholders should watch for formal validation of the proofs and any regulatory or institutional responses to allegations of improper data leverage.
