On this page
An Anthropic mathematician says he has found a proof of the Riemann hypothesis with help from the company's Claude AI model, but the claim is unverified. Levent Alpöge posted the claim on his personal website on 20 November 2026, according to Scientific American. No outside mathematician has publicly confirmed it, it has not been peer reviewed, and the Riemann hypothesis, open since 1859, should still be treated as unsolved.
This is a claim, not a result. Below is what has been reported, what has not been confirmed, and why mathematicians will wait for checking before drawing conclusions.
What has been claimed?
Scientific American reports that Alpöge, a mathematician employed by Anthropic, says that "working with Claude" he found a proof of the Riemann hypothesis. He also says the argument covers the Generalized Riemann Hypothesis, a broader version of the problem. Crypto Briefing describes the post as a brief first-person note on his website, and says he studied under Manjul Bhargava at Princeton.
OfficeChai gives more detail about the note itself:
- It is a note titled "Remark", not a paper. A human-readable write-up is promised later.
- It says the argument is formalised in Lean, split across 14 compressed archive parts.
- It expects the Extended Riemann Hypothesis could follow with more time.
- Its author warns of "suboptimal choices" and says some parts paraphrase an AI-written summary and may not make sense.
OfficeChai also reports that a footnote in the note puts the informal AI-generated draft material at more than 250,000 pages. That is the note's own figure, as relayed by one outlet, and we have not been able to check it.
What is and is not confirmed?
| Point | Status |
|---|---|
| Alpöge, an Anthropic mathematician, posted a claim on 20 November 2026 | Reported by Scientific American, Crypto Briefing and OfficeChai |
| The proof was developed with Claude | The author's own claim |
| The argument is formalised in Lean | Claimed; not independently confirmed |
| The proof is correct | Not confirmed |
| Peer review or expert endorsement | None reported |
| An official Anthropic statement | None reported by Scientific American or Crypto Briefing |
Scientific American says experts should be able to check whether the hypothesis was coded correctly and whether the proof holds, and that this is still to happen. OfficeChai goes further: nobody outside the author's circle has confirmed that the Lean archive builds or that the final theorem faithfully states the Riemann hypothesis. The note thanks several mathematicians, but being thanked is not the same as having checked the work.
What is Lean, and does it settle the question?
Lean is a proof assistant that mechanically checks every logical step of an argument. If a Lean proof compiles, its steps are valid. As Crypto Briefing puts it, a machine can confirm validity, but humans still have to confirm that the formal statement captures the right problem. A proof of a mis-stated theorem would compile and still prove nothing about the real Riemann hypothesis.
That is why the likely next steps are practical: build the project, read the final theorem statement, and check which axioms it relies on. OfficeChai notes that the note does not state the size of the formalisation, build instructions or the axioms used.
What are the Navier-Stokes claims?
OfficeChai reports that on the same day Alpöge and New York University mathematician Tristan Buckmaster posted a claimed proof that the Navier-Stokes equations can blow up without external force. If both claims held, they would be two Millennium Prize Problems in a day. OfficeChai does not report on whether that proof has been checked, and we found no second source that confirms the details, so treat it as a separate, unverified claim.
Is this the same as Anthropic's August result?
No. On 10 August 2026 Anthropic published a post saying an unreleased research version of Claude raised the known lower bound on the proportion of zeta zeros that satisfy the Riemann hypothesis from 41.6% to 67.2%. Alpöge and colleague Ralph Furman studied the paper, and outside experts Brian Conrey and Dan Goldston examined it. Claude also produced a Lean proof.
Anthropic wrote: "We don't expect that the techniques Claude used will lead to proving the Riemann hypothesis." In a Scientific American article of 12 August 2026, Oxford mathematician James Maynard said that even optimistically there was no pathway from these approaches to the actual hypothesis. The new claim therefore reverses the expectation Anthropic itself set three months ago, which is one reason for caution.
What is the Riemann hypothesis, and why does it matter?
The Riemann hypothesis, posed by Bernhard Riemann in 1859, concerns the zeros of the Riemann zeta function. It says all the nontrivial zeros lie on a single line, the critical line. Prime numbers look random, but their spread is tied to those zeros, so a proof would sharpen what mathematicians can say about how primes are distributed. Scientific American's August article notes that a $1 million prize was announced in 2000 for solving it.
Our analysis
This section is our own reading, not reporting.
Three things make this claim hard to judge quickly. First, the author's own caveats are unusually frank for a proof announcement. Second, the argument is reported to exist as a Lean archive rather than a readable paper, so human experts must work backwards from code. Third, the author works for the company whose model helped produce it, so independent checking matters more than usual.
None of that means the proof is wrong. Formal verification could, in principle, make this easier to settle than a conventional paper would. But "could" depends on outsiders building the project and confirming that the theorem statement is the real one. Until then, the sensible position is neither belief nor dismissal.
For an earlier Anthropic claim that also needed separating from verification, see our coverage of Claude's open-problem results in black hole physics.
Sources
- Scientific American: Anthropic mathematician claims to prove Riemann hypothesis
- OfficeChai: Claimed Riemann Hypothesis proof appears on Alpöge's website
- Crypto Briefing: Anthropic researcher claims Riemann hypothesis proof
- Anthropic: Riemann zeta result, 10 August 2026
- Scientific American: No, AI didn't just solve the thorniest problem in math
FAQ
Frequently asked questions
It is a conjecture from 1859 about the zeros of the Riemann zeta function. It says every nontrivial zero lies on one line, the critical line. The statement is tied to how prime numbers are distributed.
Because primes underpin number theory, and the hypothesis would give precise control over how they are spread out. Scientific American also notes a $1 million prize was announced in 2000 for a solution.
Not as far as can be confirmed. Alpöge claims a proof made with Claude, posted on 20 November 2026, but it has not been independently verified or peer reviewed. Anthropic's August result concerned a related bound, not the hypothesis itself.
If the claim holds up, it would settle a 167-year-old problem and, according to Scientific American, have sweeping consequences across number theory. It would also be a landmark for AI-assisted mathematics. That depends on expert and formal checking, which has not been reported yet.
Spotted an error? Tell us. Read how we report and correct stories in our editorial policy.
Written by
Mainstream Tech
Newsroom
News and analysis on technology, AI, business, politics and international affairs, from our newsroom in London.
More from Mainstream



