The statement and Claude's role
A mathematician working for Anthropic, Levent Alpöge, published a note on his personal website on Friday in which he claims that Claude, the San Francisco-based company's artificial intelligence software, helped prove the Riemann hypothesis. According to The Guardian, Alpöge says the model helped develop a solution over the course of about six weeks.
Why the issue matters
In the note, Alpöge describes his work with the system as a shared journey and adds that, in the final phase, Claude had surpassed him. The statement is accompanied by a reference to code written in Lean, which according to the mathematician would be capable of verifying the results. This is therefore a public claim by the author, not yet an independent confirmation.
Timing, competition and reservations
The Riemann hypothesis is a mathematical problem named after the German mathematician Bernhard Riemann, who proposed it in 1859. It concerns prime numbers and the way they are distributed along the number line; Riemann developed a formula to estimate how many prime numbers exist below a given value. The problem is one of the seven Millennium Prize Problems selected by the Clay Mathematics Institute, which has offered one million dollars for each solution.
If the proof were confirmed, it would represent a further case of AI being used for complex mathematical problems. However, the material does not make it possible to determine whether the proof is correct: verification is entrusted to the mathematicians who have begun to examine the work. The news should therefore be read as a claim to be checked, not as an established result.
Alpöge's note was published one day after Anthropic's Nasdaq listing. The author himself acknowledged that the timing is problematic, saying he would have preferred to conclude the work much earlier. The Guardian also reports that on Friday an independent Italian mathematician, Alessandro Masullo, circulated an incomplete proof of the hypothesis, which he had been working on for nearly twelve years, specifying that he had not used AI because he does not fully trust the system.
The case is part of a broader debate over the use of AI in mathematics. The Guardian recalls that in September OpenAI had claimed that a non-public version of its tools had helped solve part of another Millennium Prize problem. On that occasion, 10,000 AI systems were used at a cost of $15 million, raising concerns. Alpöge was working on that different problem for Anthropic when OpenAI made the announcement. The validity of the proof and the scientific community's judgment on the use of these tools therefore remain to be clarified.
