Anthropic mathematician claims Claude helped prove Riemann hypothesis, two groups verify
Anthropic mathematician Levent Alpöge says Claude helped him prove the Riemann hypothesis over about six weeks.
Key takeaways
- Anthropic mathematician Levent Alpöge says the company's Claude AI helped prove the Riemann hypothesis, a 19th-century problem in number theory. The Lean Focused Research Organization and the Mathlib Initiative said they verified the proof after reviewing his code, while mathematicians continue to check the work.

Key points
- Anthropic mathematician Levent Alpöge says Claude helped him prove the Riemann hypothesis over about six weeks.
- The Lean Focused Research Organization and the Mathlib Initiative said they verified his proof after reviewing his Lean code.
- The Clay Mathematics Institute offers $1 million for each Millennium prize problem solved, and a rival Italian proof is incomplete.
Levent Alpöge, a mathematician at Anthropic, posted a note on his personal website on Friday saying the company's Claude AI software had helped prove the Riemann hypothesis. He said the tool helped develop the solution over about six weeks.
The Riemann hypothesis, proposed by German mathematician Bernhard Riemann in 1859, concerns how prime numbers are distributed along the number line. It is one of seven Millennium prize problems, and the Clay Mathematics Institute has offered $1 million for each one solved.
On Saturday, the Lean Focused Research Organization and the Mathlib Initiative said they had verified the solution after reviewing the code Alpöge shared, written in a programming language called Lean. "We have full confidence that Alpöge's proof of the Riemann hypothesis is correct," the groups said.
Alpöge linked to the Lean code, which he said could verify his findings. Mathematicians then began working to check the proof.
A separate attempt came from Alessandro Masullo, an independent mathematician in Italy, who released an incomplete proof on Friday. The Guardian reported that Masullo had worked on it for almost 12 years and that he avoided AI, citing a lack of full trust in the system.
Several mathematicians are uneasy about depending on AI for problems like this. In September, OpenAI said a non-public version of its tools helped solve part of another Millennium prize problem, a result some found strange because a company spent $15 million to run 10,000 AI systems. At the time OpenAI made that announcement, Alpöge was already working for Anthropic on that separate problem. James Robinson, a University of Warwick mathematician, told The Guardian in September that big tech firms were wasting resources to show off their latest models.
Sources: Bloomberg, Financial Times, The Guardian