A mathematician employed by the artificial intelligence firm Anthropic has claimed that he successfully used the company’s Claude AI model to formulate a proof for the Riemann Hypothesis. If verified, the discovery would resolve one of the oldest and most significant open problems in mathematics, which has eluded the world’s leading scholars for over 160 years.
The Claim and Methodology
Levent Alpöge, a mathematician and researcher at Anthropic, announced the development on Friday, November 20, 2026. In a four-page note released on his personal website, Alpöge stated that he had been working with an unreleased version of Claude over a six-week period to develop the proof. He further asserted that his findings also address the generalized Riemann hypothesis.
To support his claim, Alpöge released a version of the proof written in Lean, a programming language designed to allow mathematical arguments to be formally verified by computers. “I believe I have, working with Claude, discovered a proof of the Riemann hypothesis,” Alpöge wrote, describing the result as a glimpse into the potential for future scientific breakthroughs.
The announcement arrives during a period of intense activity in AI-assisted mathematics. Earlier this year, Anthropic reported that a model had achieved progress on a related result regarding the hypothesis, and competitor firms, including OpenAI, have recently announced solutions or major advancements concerning hundreds of other long-standing mathematical problems.
Verification and Expert Scrutiny
While the prospect of a confirmed Riemann Hypothesis solution has generated significant excitement, the mathematical community remains cautious. Under the standards typically required for major mathematical prizes, such as the Clay Mathematics Institute’s Millennium Prize Problems, a proof must be published in a peer-reviewed journal and undergo a rigorous vetting process that can span several years.
The use of Lean code offers a distinct path toward verification. Because Lean acts as a formal verification tool for logical steps, experts can test whether the reasoning is soundly constructed. However, experts note that while the format allows for automated checking, the process of evaluating such a complex proof will take months, if not longer, to ensure there are no subtle errors in the logic or the underlying assumptions.
A Changing Landscape for Mathematics
The claim follows closely on the heels of Anthropic’s recent public stock market debut, adding a layer of corporate interest to the breakthrough. The rapid succession of AI-generated mathematical proofs has triggered a broader debate among academics about the nature of intellectual labor and the evolving role of machine intelligence in creative and analytical fields.
Critics and concerned scholars have noted a potential misalignment between the rapid, often opaque output of AI systems and the traditional, human-led verification standards of the scientific community. Despite these concerns, Alpöge and other proponents of AI-driven research suggest that these tools are opening new doors for human inquiry, allowing researchers to navigate complex problems that were previously considered computationally or cognitively out of reach.
As of now, Anthropic has not provided extensive commentary on the specific claim, and the mathematical world awaits the results of formal peer review to determine if the 160-year-old “Holy Grail” of mathematics has finally been solved.
