OpenAI's math claim sparks plagiarism dispute over Codex usage
OpenAI claims its internal models solved the Navier-Stokes problem, but the announcement is overshadowed by allegations from researcher Tristan Buckmaster that OpenAI scraped a Codex session and issued career threats. Buckmaster and colleague had been working on a related problem using AI assistants.
Buckmaster and Alpöge spent roughly a year on a personal, self-funded project using Claude and Codex for documentation and logic checks. They achieved verified blowup results for Euler and Boussinesq equations on August 15, 2026, confirming them via the Lean proof assistant a week later, though Buckmaster described the AI-generated proof as "AI slop."
After rumors surfaced that Anthropic had solved a major problem, Buckmaster emailed an OpenAI mathematician to clarify the work was personal. The mathematician offered compute time to avoid competing, but the announcement was overshadowed by Buckmaster's allegations that OpenAI scraped a Codex session and issued career threats.
This dispute could significantly erode trust in AI firms' handling of proprietary user data, especially among researchers who rely on these tools. If allegations of scraping and intimidation are substantiated, it may chill open collaboration in mathematics and science. The episode also highlights how AI-assisted research blurs lines between human and machine authorship, potentially reshaping academic credit and institutional competition in ways that could disadvantage independent scholars.