Anthropic employee Levent Alpöge said Friday he used the company's Claude AI chatbot to prove the Riemann hypothesis, one of mathematics' most famous unsolved problems for nearly 170 years. Alpöge said the proof took about six weeks to develop using a non-public version of Claude. The complete textual proof is over 250,000 pages long when written in Lean, a computer language used to verify mathematical proofs.

Prominent mathematician Terrance Tao said Friday that verification will take considerable time, though noted that Alpöge can help explain the work, an advantage not available for earlier AI-generated proofs.

The announcement comes days after Anthropic completed its US stock market debut. Alpöge said he messaged Anthropic CEO Dario Amodei on Sunday to ask how to announce the discovery. "The precise timing is horrible," Alpöge said, referring to the IPO. "If I could have, I'd have finished this all much earlier." He added that the timing's appearance was "obviously cynical" but false.

The claim follows recent AI announcements of solutions to major mathematical problems. OpenAI said in September that employees solved part of the Navier-Stokes equations in 88 hours using a non-public AI version. OpenAI later claimed solutions or major progress on 372 more outstanding higher mathematics problems. Those claims remain unverified.

A group of mathematics scholars criticized AI companies for announcing major mathematical solutions. "We are witnessing a general threat to intellectual work, with misalignment between the outcome of the use of AI and its initial purpose," they wrote in a group letter, saying such announcements are "detrimental to the science of mathematics."