Anthropic Mathematician Claims Claude Helped Prove Riemann Hypothesis, Math’s Most
Anthropic Mathematician Claims Claude Helped Prove Riemann Hypothesis, Math’s Most Famous Unsolved Problem

A mathematician at artificial intelligence company Anthropic said Friday that he has proven the Riemann hypothesis, widely considered the most famous unsolved problem in mathematics, with the help of the company's Claude AI model.

If confirmed, the result would be one of the most significant achievements in the history of mathematics. Mathematicians say the claim must still be independently verified.

"I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis," Levent Alpöge wrote in a statement posted on his website.

A 167-year-old problem

German mathematician Bernhard Riemann first proposed the hypothesis in 1859. It suggests that prime numbers — numbers divisible only by 1 and themselves, which appear scattered at random along the number line — actually follow a deep, hidden order.

The conjecture has become a central goal of number theory. It is one of the Millennium Prize Problems, for which the Clay Mathematics Institute has offered a $1 million reward for a correct solution.

Generations of leading mathematicians have tried and failed to prove it. Many experts spent little time working on it directly, because there was no clear path to a solution.

What Alpöge claims

Alpöge posted his announcement as a four-page note on his website. Rather than a traditional research paper, he released the proof as code written in Lean, a programming language that lets a computer check each step of a mathematical argument.

The Lean archive is massive. It contains about 175,000 source files and totals roughly 239 million bytes, according to Kingy.ai, a website that examined the release.

Alpöge said his proof also confirms the Generalized Riemann Hypothesis, a much broader version of the conjecture. Many other mathematical results depend on that version being true.

In his note, Alpöge said he steered the research toward a promising approach, and he credited Claude with rapid progress near the end. He said he hopes to simplify and explain the argument himself, and acknowledged that some of his descriptions may need correction.

The announcement did not say which version of Claude was used.

Verification under way

Mathematicians say the next step is to confirm whether the proof is correct.

Lean can check that each step of an argument follows logically from the last. But experts must also confirm that the problem itself was stated correctly in the code, and that the proof does not rely on hidden assumptions.

An independent project began checking the Lean files Friday. It confirmed its copy matched Alpöge's release, but its full computer check was still running Friday morning, according to Kingy.ai.

Scientific American reported that it should not take long for experts to determine whether the hypothesis is properly coded and whether the proof is correct.

As of Friday, the Clay Mathematics Institute still listed the Riemann hypothesis as unsolved. Under the institute's rules, a solution must be published in a qualifying outlet and generally accepted by the mathematical community, and at least two years must pass after publication, before a prize is considered.

Anthropic had not issued a public statement about the claim.

'Something like a Rosetta Stone'

If the proof holds up, it would immediately confirm a large body of mathematical results that have been proven only on the assumption that the Riemann hypothesis is true.

Hector Pasten, a mathematician at the Pontifical Catholic University of Chile, told Scientific American that a correct and understood proof would complete a dictionary between different branches of math where the connections were only partial — "something like a Rosetta Stone."

The hypothesis has no immediate practical applications. It would not, for example, change how online banking or encryption work.

AI's growing role in math

The claim caps a year in which AI systems have made rapid advances in mathematics, moving from unreliable problem-solvers to tools capable of tackling major research questions.

Alpöge has drawn attention before. In July, he posted a Claude-generated solution to the Jacobian conjecture, another long-standing problem. He was also involved in a dispute over work on the Navier–Stokes equations in September.

In October, a set of results released by rival OpenAI made significant progress on problems related to prime numbers, Scientific American reported.

Mixed emotions

Alpöge's statement acknowledged the mixed feelings many mathematicians have about AI's rapid progress in their field. He wrote that it was natural to feel a sense of loss, but said what he saw during the work left him optimistic about the future of science.

Mathematicians are expected to closely scrutinize the proof in the coming days and weeks. Until then, the claim remains unverified.