Science & Tech · United States of America
Anthropic mathematician claims Claude helped prove Riemann hypothesis
Levent Alpöge says he developed a proof with Anthropic’s AI software. The claim has not been independently verified, according to Dawn.
An Anthropic mathematician claims Claude helped prove the Riemann hypothesis, but independent confirmation is absent.
- Alpöge announced the claim on his website.
- The hypothesis dates to 1859 and concerns prime numbers.
- Alpöge says linked Lean code could verify his findings.
- Dawn reports no independent verification or outside expert review.
What's new
- Alpöge announced a claimed proof in a note on his personal website.
- He linked to Lean code that he says could verify the result.
- The Guardian reports that mathematicians are examining whether the proof can be confirmed.
Levent Alpöge, a mathematician at San Francisco-based artificial intelligence company Anthropic, has claimed on his personal website that its Claude software helped him prove the Riemann hypothesis, according to reports published on November 21, 2026. The claim remains independently unverified, Dawn reported.12
A claim awaiting verification
Dawn described Alpöge’s announcement as a four-page note and reported that outside experts had not yet reviewed the proposed proof. Alpöge expressed his conclusion as a belief: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis,” he wrote.2
According to the Guardian, Alpöge said the work with Claude took about six weeks. He also provided a link to code written in the programming language Lean, which he maintained could verify his findings. The newspaper reported that mathematicians were working to establish whether the proposed proof could be confirmed.1
A question dating to 1859
Bernhard Riemann proposed the hypothesis in 1859. It concerns how frequently prime numbers occur and how they are distributed. Dawn reported that the problem had remained unresolved for more than 160 years.12
The Guardian reported that the Clay Mathematics Institute included the Riemann hypothesis among its seven Millennium prize problems, offering $1 million for a solution to each. The New York Times framed confirmation of Alpöge’s claim as a development that could reinforce expectations about AI’s ability to tackle outstanding mathematical questions.13
Brian Conrey, a specialist on the hypothesis, responded with surprise, telling Dawn: “I really did not see this coming at all”. But his comments were not based on examining a manuscript: the newspaper said he had seen only Alpöge’s website remarks. Conrey described the proposed work as complicated and said he expected simpler approaches to emerge.2
Other approaches and competing claims
The Guardian also reported that Alessandro Masullo, an independent mathematician in Italy, had released an incomplete proof. Masullo said he had spent almost 12 years on the work and had not used AI because he did not fully trust it.1
In September, according to the Guardian, OpenAI claimed that unreleased AI tools had helped solve part of a different Millennium prize problem. The newspaper reported spending of $15 million to deploy 10,000 AI systems on that effort.1
James Robinson, a mathematician at the University of Warwick, criticised technology companies’ use of such projects to promote their models, the Guardian reported. He argued that the activity consumed fuel and, backed by billions of dollars, carried the potential for significant environmental damage.1
Why it matters
For European readers, the story includes both an independent research effort in Italy and criticism from a University of Warwick mathematician about the resources used in AI demonstrations. The central mathematical claim nevertheless remains unverified, according to Dawn.12
Videos
Related stories
Version history
- version 1 ·

