Riemann Hypothesis Proof With Claude: What Was Verified and What Wasn’t

A Riemann hypothesis proof built with AI is the biggest maths story of the year. On Friday, 20 November 2026, Anthropic mathematician Levent Alpöge said he had proved the Riemann hypothesis while working with Claude, Anthropic’s AI model. A day later, the teams behind the Lean proof checker said they had full confidence the proof is correct. Here is what happened, what has been checked, and what is still open.
Table of Contents
What Alpöge announced
Alpöge posted a four-page note on his personal website. In it he wrote: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He says the same method also proves the Generalized Riemann Hypothesis for Dirichlet L-functions.
Unusually, he did not release a normal research paper. The proof was published as code for Lean, a program that checks every logical step of a mathematical argument. According to reports, the Lean library is around 17 million lines long and was released as 14 archive files. Alpöge says an AI-written plain-language version ran to more than 250,000 pages, which he chose not to publish. He has said he plans to simplify the argument and write a human-readable account later.
Why the Riemann hypothesis matters
The Riemann hypothesis was proposed by Bernhard Riemann in 1859. It predicts where the zeros of the Riemann zeta function lie, and it is closely tied to how prime numbers are spread out. It is one of the seven Clay Millennium Prize Problems, with a $1 million prize attached, and for more than 160 years it has been the most famous unsolved problem in mathematics.
How the Riemann hypothesis proof was checked
Because the argument exists as Lean code, experts could test it by machine within hours rather than months.
- Independent compiling: MIT mathematician Andrew Sutherland compiled the proof himself on Friday and told Scientific American the chance of anything being wrong was approximately zero.
- Lean community statement: On 21 November, the Lean FRO and the Mathlib Initiative put out a joint statement saying they have full confidence the proof is correct.
- Statement checks: Two separate tools confirmed that the theorem proved matches Mathlib’s own formal statement of the Riemann hypothesis, closing the worry that the code proved a slightly different claim.
- Multiple checkers: Reports say the exported proof passed Lean’s own kernel and third-party checkers written in other languages.
| Date (2026) | What happened |
|---|---|
| 20 November | Alpöge posts his note and the Lean proof |
| 20 November | Andrew Sutherland (MIT) compiles it independently |
| 21 November | Lean FRO and Mathlib Initiative say they have full confidence it is correct |
| Still pending | Peer review, a human-readable write-up, and any Clay Prize decision |
What mathematicians are saying
Reaction has been a mix of excitement and caution. Oxford’s James Maynard told The New York Times he was 98 or 99 percent sure, according to The Next Web. Terence Tao wrote that understanding the proof will take time. Stanford’s Brian Conrad criticised releasing a proof that its own author says he does not fully understand.
That last point is the real open question. A machine check can confirm that every step is valid, but mathematicians also want to understand why it works. Alpöge has admitted he has not mastered all the surrounding mathematics and credits Claude with the final steps.
A rival proof the same day
On the same day, American Mathematical Society president Ravi Vakil shared a separate, human-written proposed proof by Alessandro Masullo. Within hours, several people reported problems in it. Unlike Alpöge’s work, it has not been formalised in Lean.
What this means for AI
If the result holds, it would be the first time an AI system helped settle one of the great open problems in mathematics. It follows earlier Claude results, including an August 2026 paper that raised the proven share of zeta zeros on the critical line to 67.2%. The announcement also landed a day after Anthropic’s record stock market debut, which some critics noted. You can read our explainer on the Anthropic IPO for that side of the story.
For now, the Riemann hypothesis proof is machine-verified but not peer-reviewed. The Clay Mathematics Institute normally requires publication and at least two years of scrutiny before awarding its prize, so no prize decision is expected soon.
Riemann hypothesis proof: the short version
In short, the Riemann hypothesis proof from Alpöge and Claude is machine-checked and backed by the Lean community, but it is not yet peer-reviewed or fully understood by humans. Expect months of expert reading before the Riemann hypothesis proof is widely accepted, and years before any prize decision.
FAQ
Has the Riemann hypothesis been proven?
A proof by Levent Alpöge, built with Claude, has been checked by machine in Lean, and the Lean FRO and Mathlib Initiative say they are fully confident it is correct. It has not yet been peer-reviewed or published in a journal.
Did AI prove the Riemann hypothesis on its own?
No. Alpöge says he worked with Claude and credits it with key steps. He is the author of the claim.
Will the $1 million Clay Prize be paid?
Not yet. Clay’s rules require publication in a qualifying outlet and a period of at least two years of scrutiny.
Where can I follow AI news like this?
See our AI news section and our guide to Gemini computer use on Mac for other big launches this week.
Sources: Scientific American, The Next Web, Startup Fortune.

