Anthropic Mathematician Claims AI Has Proved the Riemann Hypothesis, Maths' Most Famous Unsolved Problem
![]() |
| Bernhard Riemann proposed the hypothesis in 1859. It has defied mathematicians for 167 years. |
Levent Alpöge, a mathematician at Anthropic, claimed on Friday that he has proved the Riemann hypothesis — the most famous unsolved problem in mathematics — working with the AI model Claude. "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis," Alpöge wrote in a statement posted on his website.
The hypothesis, first proposed by Bernhard Riemann in 1859, claims that every non-trivial zero of the Riemann zeta function lies on a single "critical line" in the complex plane — a pattern explaining the hidden order behind the distribution of prime numbers. It is one of the Clay Mathematics Institute's seven Millennium Prize Problems, each carrying a $1 million reward — and the only one of David Hilbert's 23 problems on that list.
![]() |
| The Riemann zeta function traced along the critical line, where the hypothesis says every non-trivial zero must lie. |
Alpöge says the same method also settles the broader Generalised Riemann Hypothesis, and he expects to reach the Extended Riemann Hypothesis with more time. Rather than a traditional paper, he released the proof as computer code in Lean, a proof assistant that checks every logical step, split across 14 archive files on his website. The informal write-up ran to more than 250,000 pages, so only the Lean code was published. A timeline in the note, which Alpöge says Claude compiled, runs from 6 October to 16 November, when the final step was completed.
No outside expert has yet confirmed the claim. A Lean proof that compiles is powerful evidence, but mathematicians must still check the code proves the actual hypothesis, its definitions match the standard objects, and no hidden assumption slipped in. Alpöge is a number theorist who earned his PhD at Princeton in 2020 under Fields medallist Manjul Bhargava, won the Morgan Prize as an undergraduate, and was a Junior Fellow at Harvard before joining Anthropic.
![]() |
| The primes look random at first glance — the Riemann hypothesis says a hidden order governs them. |
The claim caps a year in which AI has stormed mathematics. Alpöge tweeted a Claude-generated solution to the Jacobian conjecture in July, was central to September's Navier-Stokes controversy involving OpenAI, and in August he and Ralph Furman showed with Claude that more than two-thirds of the zeta zeros are simple and on the critical line. In October, OpenAI released 722 AI-generated maths manuscripts. MIT mathematician Andrew Sutherland once warned: "An AI that can prove the Riemann hypothesis is not one I'd want to meet. If that happens, mathematicians having jobs will be the least of our problems."
Alpöge's own statement is elegiac as well as triumphant. "It is natural to feel that this is an end, in my case a loss of a principal guiding light," he wrote, "but what I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater." He added: "I hope the argument provides a glimpse towards the miracles that lie ahead, and shows us the joy and wonder we are about to find in the simplest things." Anthropic has not issued its own statement on the claim.
Hours after Alpöge's claim, a second proof claim emerged. Ravi Vakil, president of the American Mathematical Society, posted a draft by Stanford mathematician Alessandro Masullo claiming the Generalised Riemann Hypothesis by a wholly human route, one that had been under private review for months by Stanford's Brian Conrad. Conrad confirmed the details of the post to Scientific American but said more work is needed, adding that the paper is undergoing major revision. Neither claim has been verified by outside experts.
Sources: Scientific American (20 Nov 2026); Nacionale (20 Nov 2026); AI Parabellum (20 Nov 2026); ChainCatcher (20 Nov 2026); Wikipedia (Riemann hypothesis; Millennium Prize Problems; Levent Alpöge); qed.bot (20 Nov 2026); DataCamp (20 Nov 2026).


