Short answer, updated November 20, 2026: not confirmed. Levent Alpöge, a mathematician at Anthropic, says that he and Claude have proved the Riemann Hypothesis and the Generalized Riemann Hypothesis, and he has released the proof as Lean code. No independent expert has yet reported that the code builds and proves the right statement, and no human-readable paper exists. Until that changes, the accurate description is a claimed proof.
What exists right now
- A four-page note by Alpöge on his personal website, describing the claim, Claude's role and a 20-step timeline of the discovery.
- A Lean formalization split into 14 archive files. Alpöge says it was produced with Claude and uses no extra assumptions.
- No preprint and no journal submission. Alpöge says an informal draft produced with Claude ran past 250,000 pages, and he decided not to post it.
What has to happen before we can say "solved"
1. The code has to build
Anyone should be able to download the archive and have Lean accept every proof in it. On November 20, mathematicians reported that the archive was still compiling, a process expected to take hours.
2. The theorem has to say the right thing
Lean guarantees that a proof proves its own statement. People have to check that the statement means what we think it means: that ζ(s), its non-trivial zeros and the relevant L-functions are defined in the standard way, and that the final theorem really is RH and GRH rather than a weaker or subtly different claim.
3. No extra axioms and no gaps
Lean can list every axiom a theorem relies on. A sound result should depend only on Lean's standard foundational axioms, with no unproved placeholders (Lean's sorry) and no added assumptions.
4. Independent re-checking
For Claude's August 2026 result on zeta zeros, Anthropic reported that the Lean proof passed comparator, a validation tool from the Lean community. Independent checks of that kind are the natural next step here.
5. Human understanding
Even a fully checked formal proof leaves mathematicians wanting to know why it works. Alpöge says he intends to write the argument up properly and expects much shorter proofs to follow. Acceptance across the wider community will depend on that exposition.
Why Lean changes the timescale
Traditional peer review of a proof this ambitious could take years. A machine-checked proof shifts the hardest question from "is every step right?" to "is the statement right?". Scientific American wrote that it should not take long for experts to verify whether the hypothesis is properly coded and whether the proof is correct. That is why this claim is being taken far more seriously than the many unverified proofs posted over the decades.
A cautionary precedent
In 2018 Sir Michael Atiyah presented a claimed proof of the Riemann Hypothesis at the Heidelberg Laureate Forum, and it was not accepted. By contrast, Claude's August 2026 result was validated by two Anthropic mathematicians, examined by zeta-function specialists Brian Conrey and Dan Goldston, formalized in Lean, and later re-proved more simply by Youness Lamzouri. That is the kind of trail the November claim will need.
What about the $1 million Clay prize?
Under the Clay Mathematics Institute's rules, a proposed solution is considered only after it is published in a qualifying outlet, at least two years have passed since publication, and it has general acceptance in the mathematics community. Even if everything checks out quickly, a prize decision is years away. The rules predate machine-assisted proofs, so how Clay would assign credit for a proof made with an AI system is an open question.
Status log
- November 20, 2026: Alpöge posts his note and the Lean archive. Scientific American reports the claim. Mathematicians report the archive compiling, and discussion begins on the Lean community's Zulip chat.
We add a dated entry whenever there is verifiable news. Go back to the Riemann Hypothesis and Claude overview, read the full Riemann Hypothesis timeline, or see how Claude reached its earlier 67% result on zeta zeros.