The Lean FRO and Mathlib Initiative say they have full confidence in Levent Alpöge's Claude-assisted Riemann hypothesis proof, a 17-million-line Lean library.

The Lean and Mathlib communities, the groups behind the Lean proof-checking software, said on 21 November that they have "full confidence" in a proof of the Riemann hypothesis released a day earlier by Anthropic mathematician Levent Alpöge. Their joint statement, endorsed by the Lean FRO and the Mathlib Initiative, certifies that both the formal statement and the proof are correct. Alpöge says he built the argument with Anthropic's Claude AI over about six weeks and published it only as a Lean library of roughly 17 million lines of code. According to the statement, dozens of people compiled the code without errors, the theorem matches Mathlib's existing definition of the hypothesis, and the proof passed Lean's own kernel as well as several independent checking programs.

The Riemann hypothesis, proposed by the German mathematician Bernhard Riemann in 1859, predicts how prime numbers are spread along the number line. It is one of the seven Millennium Prize Problems, each carrying a $1 million award from the Clay Mathematics Institute. In a formal proof, software checks every logical step, so gaps that can slip past human referees are hard to hide. Separate checks published by MIT mathematician Andrew Sutherland report that all 175,187 files compiled and that the proof relies only on Lean's standard axioms, although a few runs were still in progress. Alpöge also says the method extends to a generalised version of the hypothesis for Dirichlet L-functions. The release came one day after Anthropic's Nasdaq listing, timing he himself called "horrible".

What the software cannot provide is an explanation. Alpöge writes that a version written for humans would run to more than 250,000 pages, and he expects the argument to be greatly simplified before a proper paper appears. Some mathematicians have voiced unease about leaning on AI systems and large private budgets for results of this kind, and the Lean statement itself invites the wider community to help explain the new ideas. Under the Clay institute's rules, a prize is only considered after publication in a qualifying outlet and two years of general acceptance. For now, the result is a machine-verified formal proof rather than a peer-reviewed paper, and the next step is for mathematicians to understand why it works.