Committee Confirms AI Proof of Yang-Mills Problem

A five-physicist panel found OpenAI's model cleared the first part of a Millennium Prize problem without simplifying approximations.

Updated on Nov. 12, 2026 in Mathematics

Committee Confirms AI Proof of Yang-Mills Problem

A committee of five external physicists confirmed that an OpenAI model's 1,147-page proof resolves the first part of the Yang-Mills existence problem, one of the Millennium Prize problems. The committee declared itself satisfied on November 10 after checking a Lean formalisation of roughly 1 million lines of computer code.

Why it matters

The Yang-Mills problem is closely related to models of the subatomic world used by quantum physicists, so an AI-generated proof that passes expert review marks a notable step for machine-assisted mathematics. The second part of the problem, concerning the theory's predicted mass gap, is not addressed by the proof.

The AI's proof was judged against a Lean formalisation of around 1 million lines of computer code, checked by five external physicists. The natural language version of the 1,147-page proof has not been published, and the mass-gap component is still open.

The players

OpenAI

OpenAI is an artificial-intelligence research lab that developed the unreleased model behind the proof attempt.

Edward Witten

Edward Witten is a theoretical physicist at the Institute for Advanced Study in New Jersey who helped formulate the official Yang-Mills problem statement.

The details

OpenAI announced on October 26, 2026 that its unreleased model had made progress on the problem. The physicists confirmed that the Lean formalisation matched the problem statement and that the model cleared the first part without simplifying approximations. The proof builds on techniques developed in the 1980s by Tadeusz Balaban of Rutgers University.

Timeline

  1. October 26, 2026: OpenAI announces its unreleased model's progress on the Yang-Mills problem.

  2. November 10, 2026: The five-physicist committee declares itself satisfied with the proof.

The Big Picture

The Yang-Mills problem is one of the seven Millennium Prize problems set by the Clay Mathematics Institute in 2000, none of which has been solved with a prize awarded. This AI-assisted result tests whether machine-generated proofs can clear the same expert scrutiny applied to human work on these problems.

The proof does not yet resolve the mass-gap part of the Yang-Mills problem, so the Millennium Prize has not been awarded. Readers following AI research should note that the natural language proof has not been shared for independent reading.

The takeaway

Expert verification through formal code is becoming a key test for AI-generated mathematics. Readers should watch whether the second part of the Yang-Mills problem follows, since the first part alone does not settle the prize question.

Further reading

For related background, explore more in Mathematics.

Source note: This article includes information reported by Mint.