Levent Alpöge, a mathematician at the artificial intelligence company Anthropic, claimed on Friday to have proved the Riemann hypothesis with Claude, the company's AI model, Scientific American reported.
The claim is unverified. Alpöge posted the proof in Lean, a programming system that checks each logical step, but independent experts still have to confirm that the hypothesis was coded correctly. If the riemann hypothesis solved claim holds, it would show that AI has moved from helping mathematicians to resolving one of the field's oldest open problems.
What the Riemann hypothesis asks
The Riemann hypothesis was first proposed in 1859 by the mathematician Bernhard Riemann. It claims that the prime numbers, which look random at first, are in fact determined by an intricate, meaningful order, according to Scientific American.
TechCrunch reported that the problem has been one of the major unsolved problems in mathematics for more than 150 years. Scientific American calls it the most famous open problem in mathematics and the North Star of number theory.
Progress has been rare. Scientific American reported that most top experts spent little of their time on it because there was no obvious path to a proof.
How the claimed proof is being checked
Alpöge says his proof confirms the "Generalized Riemann Hypothesis," a broader version of Riemann's original guess, Scientific American reported. The reporting so far does not settle whether the claim also covers the original hypothesis.
Scientific American reported that Lean checks the logic of each argument, but experts must still confirm that the hypothesis was coded properly and that the proof is correct. It said this should not take long. No verdict had been reported at the time of writing.
In his statement on his website, Alpöge wrote: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis." He also wrote: "It is natural to feel that this is an end, in my case a loss of a principal guiding light," according to Scientific American.
Scientific American's report does not give a calendar date for the claim.
How this follows Anthropic's earlier Riemann work
TechCrunch reported that Anthropic announced on a Monday in August that an unreleased model had made significant progress on the problem, raising the lower bound of values for which the hypothesis holds. That effort tested 650 ideas, used 60 subagents and 31 million output tokens over a day and a half, and was confirmed by two in-house mathematicians and formalized in Lean.
TechCrunch also reported that contemporary AI models still cannot solve the problem. A $1 million bounty for a working general proof remains unclaimed, it said.
The results have drawn mixed reactions. TechCrunch reported that a public declaration signed in June by prominent mathematicians raised concerns that AI could undermine core values of the field, including that proofs be attributable to specific authors. In a blog post responding to the declaration, Fields Medal winner Timothy Gowers wrote: "If we arrive at a world where mathematical theorems are no longer associated with mathematicians, maybe that won't be any more problematic than the fact that stars aren't named after astronomers and most aren't named at all."
Scientific American also reported that Alpöge drew headlines in July after he tweeted a Claude-generated solution to the Jacobian conjecture. TechCrunch separately reported that an Anthropic effort had disproved that conjecture.
Scientific American describes the story as a breaking news item that will be updated. The open questions are whether expert mathematicians confirm that the Lean formalization encodes the Riemann hypothesis correctly, and whether the $1 million bounty is claimed.