Anthropic mathematician says he and Claude proved the Riemann Hypothesis
On Friday 20 November 2026, Levent Alpöge, a mathematician at Anthropic, said he used the company's Claude AI to prove the Riemann Hypothesis, a problem open since 1859. He wrote the proof in the Lean checking language. Scientific American reports an MIT mathematician compiled it independently, but outside experts have not finished reviewing it.
Updated 21 Nov 2026

The short version
- 01Anthropic mathematician Levent Alpöge says he proved the Riemann Hypothesis working with Claude.
- 02He says the method also settles the broader Generalized Riemann Hypothesis.
- 03The proof is written in Lean, which a computer checks step by step.
- 04Scientific American says MIT's Andrew Sutherland compiled it independently without issues.
- 05AFP says the claim has not yet been reviewed by outside experts, so we label it Reported.
At a glance
| Who | Levent Alpöge, mathematician at Anthropic |
|---|---|
| Claim | A proof of the Riemann Hypothesis, made with Claude |
| Announced | Friday 20 November 2026, in a four-page note |
| Problem first posed | 1859, by Bernhard Riemann |
| Format | Lean, a computer-checkable language |
| Independent check so far | Compiled in Lean by MIT's Andrew Sutherland (Scientific American) |
| Still open | Review of the mathematics by outside experts |
| Status | Reported |
What did the Anthropic mathematician claim?
AFP reports that Levent Alpöge, a mathematician at Anthropic, said on Friday that he had used Claude to prove the Riemann Hypothesis. In a four-page note he wrote: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis."
He also says the method settles a broader version, the Generalized Riemann Hypothesis. Scientific American describes the problem as the most famous open problem in mathematics, first proposed in 1859.

What is the Riemann Hypothesis?
It is a guess about prime numbers, the numbers divisible only by themselves and 1. Scientific American says it claims the pattern of the primes along the number line, which looks random, follows a hidden order.
The Clay Mathematics Institute offers a $1 million prize for a solution. According to AFP, its rules require publication in a peer-reviewed journal and two years of scrutiny, so no prize has been awarded.
How has the proof been checked so far?
The proof is written in Lean, a programming language that lets a computer check each step of an argument. Scientific American reports that MIT mathematician Andrew Sutherland and others compiled it independently in Lean without problems.
Sutherland is quoted as saying the chance that anything is fishy "is approximately zero". A further check, called a comparator run, was still awaited when Scientific American wrote. It confirms the code does not redefine the symbols in the statement.
AFP, in its report, says the proof has not yet been reviewed by outside experts, and Alpöge says he has not yet mastered all the mathematics involved.
The Lean proof has been compiled and independently rechecked by the Lean kernel with only the three standard axioms.
Andrew Sutherland, MIT mathematician, to Scientific American
How does this relate to Anthropic's earlier Riemann work?
In August 2026 Anthropic published research saying an unreleased Claude had raised the known share of zeros on the critical line from 41.6% to 67.2%. Anthropic said then that it did not expect those techniques to prove the hypothesis.
Friday's claim is a different, much bigger one. It is Alpöge's own statement, and AFP says Anthropic did not immediately respond to a request for comment.
Why does it matter?
If the proof holds, an AI helped settle the most famous open problem in mathematics. Scientific American says it would not change banking or the internet right away, but many results that assume the hypothesis could become proven facts.
The claim also lands a day after Anthropic's stock market debut. Alpöge himself acknowledged the timing could be viewed cynically, AFP reports.
What does it mean for you?
For most people, nothing changes this week. This is a maths result, not a product, and you cannot try it in Claude today.
Who should watch: anyone following whether AI can do original research. Who can ignore it for now: anyone waiting for a new feature. It is a sign of what labs are aiming at, in the same month Anthropic launched [Claude Fable 5.2 and Mythos 5.2](/news/claude-fable-5-2-mythos-5-2/), its newest models.
How to read it: a Lean proof that compiles is strong evidence, but mathematicians still want to understand and check what it says. Treat any headline that says the problem is solved as early.
One thing to do now: wait for the expert checks before sharing it as settled. We also covered [Anthropic's IPO pricing](/news/anthropic-ipo-priced-83-dollars-ant/), and what changed since is that the company is now trading.
What's next?
Experts will keep checking the Lean code and the statement it proves. Alpöge says he plans to write the argument up properly, and the Clay Institute's rules mean any prize would take years. We will update this page when Anthropic or the experts say more.
Questions people ask
Has the Riemann Hypothesis been officially solved?
No. It is a claim by one Anthropic mathematician. The Clay Institute requires a peer-reviewed publication and two years of scrutiny.
What is Lean?
Lean is a programming language and proof checker. If a proof written in it compiles, the computer has checked each step follows the rules.
Did Claude do it alone?
Alpöge says he worked with Claude and wrote that Claude helped him discover the proof. We have no further detail on who did which part.
Does this affect internet security?
Scientific American says it has no immediate real-world applications and will not upend banking or the internet.
Can I use this in Claude today?
We have seen no announcement of a product. Anthropic's August research used an unreleased version of Claude.
Sources
- PrimaryAFP via France 24: Anthropic researcher says Claude solved math's greatest riddle (20 Nov 2026)
- CoverageScientific American: Anthropic mathematician claims to prove Riemann hypothesis (20 Nov 2026)
- CoverageScientific American: The AI Riemann hypothesis proof is almost certainly true, independent mathematicians attest (20 Nov 2026)
- CoverageAnthropic: Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function (10 Aug 2026)
Why "Reported"? We label a story "Reported" when only the media have it so far. Spotted an error? Email hello@openloopnews.com and we will correct it.



