Overview
- OpenAI says an internal multi-agent model trained from late August produced a solution to the Navier–Stokes Millennium Prize problem and dozens of other long‑standing results.
- The company released machine-checkable proof files and says it will work with an independent advisory panel of mathematicians to guide how results are presented and released.
- Leading researchers remain skeptical about the panel’s scope and whether OpenAI will follow academic norms for attribution and careful peer review.
- Mathematician Andreas Thom and others have publicly raised concerns that prior ChatGPT interactions or unpublished drafts may have influenced the model, prompting renewed demands for training-data clarity.
- Formal verification and review are ongoing, with the Clay Mathematics Institute and community checks seen as crucial steps before the field accepts the AI-produced results and before norms for release are rewritten.