OpenAI’s unreleased AI model produced a possible proof for the Navier–Stokes equations, one of mathematics’ Millennium Prize Problems, just four days after being tasked on September 1. The Navier–Stokes equations describe fluid motion and have puzzled mathematicians for decades. Australian mathematician Tristan Buckmaster, who has spent years researching this problem, found the AI’s rapid progress remarkable but raised concerns about data usage, according to inc42.com.
Buckmaster and his collaborator Levent Alpöge had shared drafts of their unpublished research with OpenAI’s coding tool, Codex. Buckmaster worried that the AI might have incorporated his unpublished work into its solution, but OpenAI denied this, stating that while user data interactions may be used to improve models depending on settings, they did not influence the AI’s output in this case. The episode highlights tensions between researchers and AI developers over data privacy and intellectual property.
The Navier–Stokes problem is one of seven Millennium Prize Problems, each carrying a $1 million reward for a correct solution. The AI’s ability to produce a potential proof in days contrasts with the years human mathematicians have spent on it, underscoring AI’s growing role in complex scientific research. This case also raises questions about how AI models are trained and the transparency of their data sources, issues critical to the future of AI-assisted discovery.
Buckmaster’s experience exemplifies the challenges researchers face when using AI tools that might access sensitive or unpublished data. OpenAI’s acknowledgment that user data can be used to improve models depending on settings was confirmed in their public statements. The incident occurred in early September and has sparked ongoing discussions about AI ethics and data governance in scientific communities.