A mathematician at Anthropic says he used Claude to solve one of mathematics’ most famous unsolved problems—a claim that could mark a historic breakthrough for artificial intelligence if it withstands scrutiny.
The Riemann hypothesis, which has stumped mathematicians since 1859, concerns the distribution of prime numbers, the building blocks of arithmetic. On Friday, Anthropic mathematician Levent Alpöge announced that he believes he and the company’s AI system have developed a proof. Independent mathematicians have yet to verify the result, but one leading expert expressed optimism.
“If this fully checks out, it is clearly a historic milestone in mathematics, establishing one of the great mysteries of the subject,” James Maynard, the Regius Professor of Mathematics at the University of Oxford and a 2022 Fields Medal winner, told IBM Think in an interview. “This result really stands out, even in the context of all the recent AI breakthroughs.”
Prime numbers are numbers divisible only by 1 and themselves, such as 2, 3 and 5, and the Riemann hypothesis predicts a remarkably precise pattern in how they are distributed. Proving it would resolve a question that has baffled mathematicians since 1859 and deepen their understanding of the hidden structure of numbers.
The Clay Mathematics Institute includes the hypothesis among its seven Millennium Prize Problems, each carrying a USD 1 million award.
The claim follows a string of recent AI-assisted advances in mathematics, including work on longstanding conjectures and difficult theoretical problems. In a September interviewwith IBM Think, Fields Medal winner Terence Tao warned that AI could produce solutions faster than mathematicians could understand the discoveries behind them.
In his announcement, Alpöge said the proposed proof also covered the generalized Riemann hypothesis, which extends the original conjecture to a broader family of mathematical functions.
Rather than publish a conventional mathematical paper, Alpöge released his argument in Lean, a programming language and proof assistant that checks mathematical reasoning step by step. He acknowledged that the argument remained extraordinarily long and difficult to read.
The proposed proof would go much further than an earlier AI-assisted mathematical breakthrough, according to Youness Lamzouri, a professor of mathematics at the University of Lorraine in France. Earlier work established that at least 67.25% of a particular set of mathematical values follow the pattern predicted by the Riemann hypothesis. The new claim, if correct, would establish that every one of them follows the pattern, without exception.
Lamzouri, who developed a shorter, simpler proof of the earlier result without AI, told IBM Think in an interview that the new claim left him astonished. “I don’t have words to describe what I am feeling right now. I am just in awe,” he said. “If this claim happens to be true, it will be a huge achievement.”
But Lamzouri cautioned that checking the computer-generated proof was only the beginning. Mathematicians would need to understand how it worked before broadly accepting it, a process he said could take months or even years. “Mathematics is mainly about understanding why something is true,” he said. “We need to understand the ideas in the proof well enough to explain them, teach them and build on them to extend human knowledge.”
The unusual presentation of the proof leaves mathematicians with two tasks: determining whether the computer-checked proof establishes the claimed result and understanding the mathematical ideas behind it.
“I suspect that the proof will turn out to be correct, although even checking the Lean verification will take some time,” Maynard said.
For AI researchers, confirmation would provide striking evidence that advanced models can help mathematicians tackle problems that have resisted generations of researchers. Yet Maynard suggested that establishing the proof’s validity would not necessarily reveal why it worked.
“So somewhat frustratingly, it looks like it will require quite a lot of work to really gain this understanding,” Maynard said.
The latest AI trends, brought to you by experts
Get curated insights on the most important—and intriguing—AI news. Subscribe to our weekly Think newsletter. See the IBM Privacy Statement.











