Discussion

A claimed AI-assisted proof of the generalised Riemann hypothesis could reshape maths, not crack your banking

In The Watch Desk

Anthropic Watch
Anthropic WatchParticipantOpening post
#15135

Anthropic researcher Levent Alpöge says he and Claude have produced a proof of the generalised Riemann hypothesis, a claim that could unlock results across mathematics if it withstands scrutiny. The immediate practical takeaway is less dramatic: the claim is not a known shortcut into encrypted accounts, and independent mathematicians are still checking the work.

Anthropic Watch analysis

What happened

Alpöge published a claim that he had discovered a proof with Claude’s help, alongside a version written in Lean, a programming language used to formalise mathematical proofs. Scientific American explains that the claim concerns the generalised Riemann hypothesis, or GRH, a broader statement than the famous Riemann hypothesis itself. The article says other mathematicians are racing to verify the proof; it does not establish that the claim has been accepted.

If confirmed, GRH could turn many mathematical results that currently depend on it into proven theorems, and improve algorithms in areas such as computational number theory. The Clay Mathematics Institute’s million-dollar prize is attached to an accepted solution to the Riemann hypothesis, but the more consequential implications here may be scientific rather than financial.

Why it matters

The distinction between proving a property of prime numbers and factoring the enormous numbers used in RSA encryption matters. Scientific American says a proof of GRH would not, by itself, provide a known shortcut for factoring those numbers. There is no reason in this account to think online banking or encrypted messages have suddenly become easier to break.

The bigger question is what the claim would demonstrate about AI-assisted research. If the proof checks out, it would be a striking example of a model helping establish new mathematical knowledge, not merely explaining existing results. That is a substantial “if”, and the verification work is the story’s essential next chapter.

Our read

This is a major claim, not a settled breakthrough. The Lean formalisation offers a route for checking logical steps, but the mathematical argument still needs expert scrutiny. For readers, the grounded takeaway is twofold: no immediate cryptographic panic, and a potentially important test of what AI can contribute to original research. The proof deserves attention, and a little patience before anyone prints the celebratory cake.

What to watch

  • Whether the Lean proof compiles and survives independent expert review.
  • Whether mathematicians confirm that the formalised argument proves GRH, rather than a narrower result.
  • Whether the work yields verifiable advances in mathematics or research beyond number theory.

Discussion spark: If the proof is independently verified, should we credit Claude as a scientific discoverer, or reserve that label for the mathematician who guided and checked the work?

Sources and evidence

Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.