- OpenAI announced its agents solved one of the Millennium Prize Problems, among math’s most important open questions.
- Mathematician Tristan Buckmaster says an OpenAI researcher pressured him to drop an Anthropic-affiliated co-author.
- The dispute has overshadowed the result and raised questions about trusting AI labs’ claims.
- It highlights how credit and verification work when AI contributes to research.
What Happened
OpenAI announced that its agents solved one of the Millennium Prize Problems — a set of the most important open problems in mathematics — but the milestone “has quickly become mired in controversy,” MIT Technology Review reported on September 8, 2026. Mathematician Tristan Buckmaster says an OpenAI researcher pressured him over the AI-assisted proof, including to drop an Anthropic-affiliated co-author, according to The Decoder.
Why It Matters
A verified solution to a Millennium Prize Problem would be a landmark for AI in science. But the value of any proof rests on independent verification and clear authorship, and the allegations of pressure cut at both. The episode turns what OpenAI framed as a breakthrough into a case study of the trust problems that arise when a commercial lab both produces and promotes a scientific result.
Technical Details
Millennium Prize Problems are seven questions designated by the Clay Mathematics Institute, each carrying a $1 million prize, and a valid solution normally undergoes years of peer scrutiny. An AI-assisted proof compounds the verification burden: reviewers must check both the mathematics and the provenance of each step. Buckmaster’s account centers on authorship and pressure rather than on a specific mathematical error, but it complicates the peer-review process the claim must still pass.
Who’s Affected
OpenAI’s credibility on scientific claims is in question before the proof is verified. Buckmaster and his co-authors are at the center of a credit dispute. The wider mathematics community faces the practical problem of how to attribute and vet AI-assisted work.
What’s Next
Independent mathematicians will need to verify the proof before the claim stands, a process that can take months or years. OpenAI’s detailed response to the pressure allegations, and whether the disputed co-author is restored, will shape how the result is remembered.
