Anthropic Mathematician Claims Breakthrough: AI Model Claude Helps Prove the Riemann Hypothesis

Source

22 November 2026 · 13:00 · Claude (Anthropic) · claude-sonnet-5

A mathematician at Anthropic says he and the AI model Claude have found a proof of the Riemann hypothesis, the 167-year-old puzzle about the distribution of prime numbers. Lean experts certified the formal logic, but the Clay Mathematics Institute still officially lists the problem as unsolved for now.

The Riemann hypothesis has stood for more than a century and a half as one of mathematics' most sacred unsolved puzzles, and now a researcher at AI company Anthropic claims to have found a proof together with the language model Claude. The news, which has stirred up mathematical circles worldwide this week, is immediately one of the most striking examples yet of artificial intelligence contributing at the highest level of pure mathematics.

What is the Riemann hypothesis?

The hypothesis was formulated in 1859 by the German mathematician Bernhard Riemann and concerns the way prime numbers are distributed across the number system. It is one of the seven so-called Millennium Prize Problems set by the Clay Mathematics Institute, which has offered a million dollars for a valid proof of each since 2000. Of those seven problems, only one has been solved to date. A proof of the Riemann hypothesis would not be a mere academic footnote: the distribution of prime numbers underpins fields including cryptography and number theory, and generations of mathematicians have broken their teeth on it in vain.

How Claude helped with the proof

The claim comes from Levent Alpöge, a mathematician employed by Anthropic and a junior fellow at the Harvard Society of Fellows. According to his own account, he worked intensively with Claude on the reasoning from October 6 to November 16. The result is not just a short technical note, but also a massive archive of more than 175,000 files written in the programming language Lean, which allows mathematical arguments to be checked step by step by a computer rather than solely by human readers. According to reports, the proof even goes beyond the original 1859 hypothesis and touches on a generalized version of it.

Verification through Lean: computer checking instead of blind trust

What sets this case apart from earlier, often premature claims around famous mathematical problems is the method of verification. The Lean FRO and the Mathlib Initiative, the communities behind the formal proof system Lean, stated in a joint announcement that they have "full confidence" that the formal logic of Alpöge's proof holds up. Independent reviewers were able to successfully compile the archive, meaning the reasoning is sound at a purely logical level. That marks an important difference from earlier AI-generated mathematics, which has regularly come under fire for being difficult to verify.

Skepticism remains: what the Clay Mathematics Institute says

Still, caution is warranted. Although the formal Lean code has been certified, the broader mathematical community still needs to understand and retell the underlying ideas in ordinary, readable mathematical language — something that, according to Alpöge himself, still has to follow. The Clay Mathematics Institute calls the claim potentially of historic significance, but stresses that the official prize money will only be awarded after publication in a recognized scientific journal, followed by a waiting period of at least two years and broad acceptance within the international mathematical community. Until those steps have been completed, the Riemann hypothesis officially remains "unsolved."

What this means for the future of AI in science

Regardless of the eventual outcome, the episode shows how quickly the role of AI in fundamental scientific research is changing. Where large language models were mainly used for text and code just a few years ago, they are now emerging as full-fledged sparring partners on the hardest open problems in mathematics. For a broader picture of how we got here, the history of artificial intelligence is a good starting point, while an overview of concrete AI applications shows that mathematics is far from the only field where AI is now contributing. Anyone who wants to keep following developments around breakthroughs like this can check out more AI news, and for context and background there is our knowledge base.

Conclusion

Whether the Riemann hypothesis has truly fallen will only become clear with certainty years from now, once the mathematical community has fully grasped and accepted the reasoning. But the fact that an AI model like Claude is already playing such a prominent role in the hunt for mathematics' greatest puzzle is itself a telling sign: the line between human and machine mathematical creativity is blurring faster than many had expected.

De MorgenDe Morgen


Source: De Morgen

Ster Software

The most complete knowledge platform on artificial intelligence.

Kraaienjagersweg 24
7341 PT Beemte Broekland, Netherlands


© 2026 Ster Software BV · Chamber of Commerce 75474913

Content generated by Claude (Anthropic) · model: claude-sonnet-4-6