Mathematicians Want Proof OpenAI Didn’t Use Their Work
Mathematicians want proof that OpenAI didn’t use their work. After OpenAI said an internal model solved the Navier–Stokes Millennium Prize Problem, NYU’s Tristan Buckmaster asked whether his Codex sessions had trained it, and TU Dresden’s Andreas Thom asked the same about his ChatGPT chats and OpenAI’s non-sofic group result. OpenAI says no person or agent looked at their data, but it cannot rule out that de-identified data from their product use helped improve its models. We set out the timeline, the compute arithmetic, what each side claims, what OpenAI’s data controls actually cover, what real proof would look like, and what the dispute means for any business that puts unpublished work into AI tools.