Anthropic employee claims AI helped prove the Riemann hypothesis
A mathematician claims to have used a private version of Claude to produce a 250,000-line proof for one of the most famous unsolved problems in history.

The short version
- The claim involves the Riemann hypothesis, a problem concerning the order of prime numbers on the number line.
- The proof was created over six weeks using a non-public version of Anthropic's Claude chatbot.
- The full proof is over 250,000 lines long and written in Lean, a language used to verify mathematical logic.
- Prominent mathematician Terrance Tao noted that verifying the proof will be a long process for the community.
Why it matters
If verified, this would be one of the most significant mathematical breakthroughs in history. It also highlights a growing trend of AI companies using their tools to tackle complex scientific problems that have eluded humans for centuries.
A 170-year-old mystery
A mathematician and Anthropic employee, Levent Alpöge, has claimed to have solved the Riemann hypothesis using the company's AI chatbot. This specific problem is widely considered to be the single most famous unsolved mystery in the field of mathematics. It has remained a challenge for scholars for nearly 170 years.
The hypothesis centers on prime numbers, which are the fundamental building blocks of higher mathematics. Specifically, the theory suggests that prime numbers follow a determined order along the number line. Because prime numbers are so central to the field, any confirmed breakthrough regarding their behavior would have a broad impact on science.
Alpöge described the potential discovery as "one dawn across science of something far greater." He worked on the proof over a period of about six weeks. He used a non-public version of Anthropic's Claude to generate the detailed results.
How the proof was built
The proof was not written in standard prose but in a computer language called Lean. Lean is a specialized tool used by researchers to confirm the absolute correctness of mathematical proofs. By using this language, the author ensures that the logic can be checked by a computer system.
The resulting document is massive, totaling over 250,000 lines of code. This scale of work suggests a high level of complexity that would be difficult for a human to produce alone in such a short timeframe. Alpöge reached out to Anthropic CEO Dario Amodei to discuss how to officially announce the discovery.
The timing of the announcement was complicated by Anthropic's recent debut on the US stock market. Alpöge expressed a desire to have finished the work much earlier but stated he felt he could not wait any longer to share it. He noted that while some might see the timing as cynical, he felt a personal responsibility to move forward.
The verification process
While the claim is significant, the mathematical community emphasizes that the proof is not yet verified. Prominent mathematician Terrance Tao stated that the process of checking the work will take some time. He noted that because there is a human author to communicate with, the task might be slightly easier than verifying a purely autonomous AI output.
This verification is a rigorous process where experts must examine the logic of the 250,000 lines. Because the proof is so long, it will likely take months, if not longer, for the global community to reach a consensus. Until then, the claim remains a possibility rather than a confirmed fact.
This situation mirrors other recent events where AI firms have claimed solutions to difficult problems. For example, OpenAI previously claimed that employees used non-public tools to solve parts of the Navier-Stokes equations. Those solutions also faced criticism and have yet to be fully confirmed by the academic community.
Controversy in the field
The use of AI to solve high-level math has sparked a debate among scholars about the integrity of scientific work. Some mathematicians argue that AI companies boasting about these solutions is detrimental to the science of mathematics. They worry that the focus on commercial products may misalign with the original purpose of mathematical research.
Critics in the community have expressed concerns about a general threat to intellectual work. They argue that using AI to bypass the traditional struggle of discovery might change how we value human achievement. These scholars have even issued group letters expressing their concerns about this shift in the academic landscape.
Despite these concerns, the pace of AI progress in mathematics is accelerating. Following the Navier-Stokes claims, OpenAI reported progress on 372 other outstanding higher math problems. These claims will also require months of expert review to determine if the AI-assisted results are actually valid.
What to watch
- Whether the 250,000-line Lean proof is eventually verified by the math community.
- How other AI labs respond to the use of private models for scientific discovery.
Try it
Look up the 'Riemann hypothesis' on a math website to see why it is considered the most famous unsolved problem in the field.
Sources and links
- Officialmaths, external anthropic.com
- Officialreleased, external openai.com
- ReportingBBC: the article we sourced
First reported by BBC. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.