OpenAI's claimed AI proof of the Navier-Stokes millennium problem ignites scooping and data-use allegations

OpenAI says an unreleased model running ~10,000 agents solved the $1M Navier-Stokes Millennium Problem in 88 hours; NYU mathematician Tristan Buckmaster alleges OpenAI scooped and pressured him over joint work with Anthropic's Levent Alpöge, and OpenAI admits…

OpenAI announced that an unreleased internal model, running a swarm of roughly 10,000 AI agents for 88 hours, produced a solution to the Navier-Stokes Millennium Prize Problem — a $1 million Clay Mathematics Institute challenge unsolved by humans for nearly 90 years — and published a blog post defending the work, with Sam Altman publicly backing the team. NYU professor Tristan Buckmaster and his co-author Levent Alpöge, a researcher at Anthropic, allege OpenAI scooped their joint work. Buckmaster says OpenAI pressured him after his research leaked, that Alpöge was sidelined, and that OpenAI may have trained on drafts he entered into its systems; he questioned whether OpenAI accessed his Codex sessions. OpenAI denies using specific user data but acknowledges it cannot rule out that de-identified product-usage data helped improve its models. The company says it will not claim the $1 million Clay bounty. Alpöge says he would have welcomed cooperation with the lab, and Terence Tao warned the episode could discourage researchers from sharing work, reversing centuries of open science, cautioning that rumors alone can now trigger AI-powered races that flatten original research projects. Critics say the rushed, reportedly million-dollar effort violates academic norms of trust and openness and could chill collaboration in mathematics.

• OpenAI says an unreleased internal model solved the Navier-Stokes Millennium Prize Problem in 88 hours using roughly 10,000 AI agents (The Verge).
• The Navier-Stokes problem is a $1 million Clay Mathematics Institute challenge unsolved by humans for nearly 90 years; OpenAI says it will not claim the bounty (The Verge).
• NYU professor Tristan Buckmaster and co-author Levent Alpöge, a researcher at Anthropic, allege OpenAI scooped their joint work.
• Buckmaster alleges OpenAI pressured him after his research leaked, sidelined Alpöge, and may have trained on drafts he submitted; he questioned whether OpenAI accessed his Codex sessions.
• OpenAI denies using specific user data but acknowledges it cannot rule out that de-identified data from product usage helped improve its models.
• Sam Altman defended the team's conduct; Alpöge says he would have welcomed lab cooperation (The Decoder).
• Terence Tao warns the episode could discourage researchers from sharing work and that rumors alone can trigger AI-powered races that flatten original research projects (The Decoder).
• Critics say the rushed, reportedly million-dollar effort violates academic norms around trust and openness (The Verge).