Developing

Anthropic Researcher Claims Claude Proved Riemann Hypothesis

By The Indus Pulse Editorial Team3 min read

Anthropic mathematician Levent Alpoge released a four-page note on November 20, 2026, claiming that he and the company's Claude AI model discovered a proof of the Riemann Hypothesis, a puzzle that has eluded mathematicians for more than 160 years, according to The Hindu. Bernhard Riemann introduced the prime-number conjecture in 1859, and the Clay Mathematics Institute included it in 2000 among seven Millennium Prize Problems, The Hindu reported. Only the Poincare conjecture has been officially solved among those seven problems, and claiming the $1 million prize requires peer-reviewed journal publication followed by two years of scrutiny, according to The Hindu and The Verge. Alpoge's claimed proof has not yet been independently verified or reviewed by outside experts, The Hindu reported.

Scope of the Claimed Proof

Alpoge wrote that the method also resolves the broader Generalised Riemann Hypothesis, and The Verge quoted his statement that he expects it to extend further with more time. He added that he and Claude produced a version of the proof in Lean, a programming language used to verify mathematical logic by computer, and that he is releasing only that version for now, The Hindu reported. Alpoge said he texted Anthropic chief executive Dario Amodei late Sunday night about the result and held off on announcing it until after Anthropic's record-setting Thursday initial public offering, The Verge reported. He also wrote that he had not yet mastered all the mathematics involved, while Anthropic did not immediately respond to requests for comment from AFP or The Verge, according to The Hindu and The Verge. In August, an unreleased research version of Claude had raised the proven lower bound for the fraction of zeros of the Riemann zeta function satisfying the Riemann hypothesis from 41.6% to 67.2%, according to Anthropic.

A full natural-language write-up that Levent Alpoge generated for the claimed proof ran past 250,000 pages, according to DataCamp. On November 20, 2026, MIT mathematician Andrew Sutherland told Scientific American that he had independently checked Levent Alpoge's result in Lean, but a list of the axioms it relies on and a check against the standard statement have not been reported, according to Nacionale.

Previous AI Math Claims and Rivalry

Anthropic emphasized at the time that the August result did not prove the conjecture itself, The Hindu reported. Earlier in July, Alpoge said he had solved the Jacobean conjecture using Claude, according to The Verge. The claim follows a September announcement by rival OpenAI that it spent millions of dollars and deployed 10,000 agents to solve the Navier-Stokes Millennium Prize problem, The Verge reported. New York University mathematics professor Tristan Buckmaster accused OpenAI of lifting his and Alpoge's unpublished route via Codex and Claude, an allegation OpenAI denied, according to The Verge. Under the revised Millennium Prize rules adopted on September 26, 2018, the Clay Mathematics Institute does not accept direct submissions of proposed solutions and will not offer any further guidance or advice, according to Clay Mathematics Institute. Grigori Perelman solved the Poincaré Conjecture in 2003, according to DataCamp.

Spotted an error in this article? Tell us and we will check it.