OpenAI claims its AI agents have cracked the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems carrying a $1 million reward. This breakthrough suggests that fluid dynamics equations can break down under certain conditions, predicting impossible states like infinite velocity. Yet the announcement is drowning in controversy rather than celebration. Accusations are flying that OpenAI leveraged unpublished work by NYU mathematician Tristan Buckmaster and Anthropic employee Levent Alpöge without proper credit. The company denies accessing their private transcripts but admits a rumor about their progress inspired the team to pursue the problem.
Buckmaster posted a detailed account of his interactions with OpenAI staff on Mastodon. He alleges they offered a stark choice: post their partial proof now while OpenAI released the full solution the next day, or collaborate on a paper that explicitly excluded Alpöge due to his employment at rival Anthropic. This ultimatum highlights a growing friction between traditional academic collaboration and the secretive, resource-heavy operations of frontier AI labs. While OpenAI’s internal model reportedly outperforms the recently released Astra model, the ethical cloud hanging over the achievement is thick. The company stated it will not claim the cash prize, but the reputational stakes are far higher.
The mathematical community is left questioning how independent these solutions truly were. Both teams utilized an approach pioneered by Diego Córdoba and Luis Martínez-Zoroa. Javier Gómez-Serrano, a mathematics professor at Brown University, noted this method was already seen as promising. Independent discovery is possible, but the timing and specific tactical moves raise eyebrows. If AI models trained on private user interactions inadvertently absorbed research insights, the foundation of intellectual property in mathematics could crumble. We are entering an era where solving the hardest problems requires compute power only a few corporations possess.
This incident signals a potential shift in who drives mathematical progress. Human mathematicians may find themselves sidelined or forced into uneasy partnerships with tech giants that operate outside academic norms. The open exchange of ideas, which has long undergirded scientific advancement, faces a new threat from proprietary black boxes. Whether OpenAI’s agents truly worked in isolation or stood on the shoulders of uncredited peers, the precedent set here matters. It challenges us to redefine authorship and credit in an age where artificial intelligence blurs the line between tool and collaborator.
