Separating AI hype from reality.

Major Outlets Amplify Unverified Claim That Anthropic Mathematician Solved Riemann Hypothesis With AI

November 22, 2026 · Anthropic · Security

On Friday, November 20, Anthropic mathematician Levent Alpöge posted a four-page note on his personal website declaring that he and the company's AI model Claude had proved the Riemann hypothesis—the 167-year-old conjecture about the distribution of prime numbers that carries a $1 million Millennium Prize. By Saturday, CNN, Yahoo, Al Jazeera, Deutsche Welle, ABC News, and multiple NPR affiliates had published stories reporting the claim. The proof remains unverified. Alpöge himself acknowledges he has not mastered all the mathematics involved.

The claim and its release

Alpöge's statement says he believes he has "working with Claude, discovered a proof of the Riemann Hypothesis" that "easily extends much further, resolving the so-called Generalized Riemann Hypothesis." He released the argument not as a conventional paper but as a 17-million-line Lean formalization—roughly 250,000 pages of computer-checked code—hosted on his site in 14 archive parts totaling 239 megabytes [1][2][13].

Lean is a proof assistant that verifies each logical step. The top-level theorem statements compiled cleanly against Lean's standard mathematics library (Mathlib), and an independent verification project confirmed the downloaded archive matched Alpöge's mirror [2]. But the full automated replay of the entire proof was still running as of Friday [2], and mathematicians are still translating the code to understand it [6].

What mathematicians are checking

Experts stress three hurdles before the problem can be considered solved [2]:

  1. Exact target: That the Lean theorem statement is genuinely the standard Riemann hypothesis, not a subtly altered version.
  2. Faithful definitions: That all mathematical objects in the code accurately reflect standard definitions.
  3. No hidden assumptions: That no unproven shortcuts or unintended axioms were slipped in.

Kevin Buzzard, professor of pure mathematics at Imperial College London, told Al Jazeera that at the minute, the proof is just millions of lines of gobbledygook written in a strange programming language, and that right now, as we speak, there are people all over the world who are translating parts of this programming language back into English [6]. He added that nobody yet understands what the proof actually shows [6].

Alpöge's own caveats

Alpöge wrote that he is "presenting the world with a complete argument before having mastered a massive neighbourhood around its mathematics" [4][9]. He noted the proof's "precise timing is horrible"—it came late Sunday night, and he messaged Anthropic CEO Dario Amodei about "my ability to announce" as the company prepared for its IPO [13]. Anthropic debuted on the Nasdaq Thursday, November 19, in the largest IPO in history, raising $81 billion at a $1.5 trillion valuation [4][5]. Alpöge's announcement came the next day.

Anthropic did not respond to Associated Press requests for comment on Alpöge's claim [4][9][12].

Institutional response

Hours after the announcement, the Clay Mathematics Institute still listed the Riemann hypothesis as unsolved on its website. President Martin Bridson called the announcement "of historic significance" but added: "The journey from today's announcement to a place of real understanding will be long and difficult" [4][9]. Under Clay's rules, a solution must be published in a qualifying peer-reviewed outlet, survive at least two years, and achieve general acceptance before the prize is awarded [2].

Competing claims and context

The same day, NYU mathematician Tristan Buckmaster and Alpöge posted a separate 79-page paper claiming a breakthrough on another Millennium Prize problem—the Navier-Stokes equations—though its supporting numerical certificate has not been released [2][7]. In September, OpenAI claimed its technology helped solve Navier-Stokes, a result still unconfirmed [3][8].

Adding further confusion, Stanford postdoctoral researcher Alessandro Masullo posted a paper Friday claiming he had also proved the Riemann hypothesis—without AI—after 12 years of work, saying he was "blindsided" by Alpöge's announcement and rushed his release [5].

Outlet coverage vs. verification status

The major-outlet stories largely led with the claim and its historic framing. CNN's headline read: Anthropic worker claims to have solved a Holy Grail of math problems with the help of AI [4]. Yahoo's headline: Anthropic worker claims to have solved famed 167-year-old maths problem with help of AI [5]. Al Jazeera asked: Riemann hypothesis: Has the world's most famous maths problem been cracked? [6]. DW's headline: Anthropic researcher uses AI to solve math's greatest riddle [8]. ABC News and multiple NPR affiliates (WHRO, KANW, WAMC) ran the same AP wire story [9][10][11][12].

Most outlets included caveats—usually several paragraphs in—that the proof is unverified, that Alpöge admits gaps in his own understanding, and that the Clay Institute still lists the problem as unsolved. But the headlines and leads presented the claim as a live breakthrough rather than an unchecked assertion posted to a personal website the day after the author's employer went public.

What happens next

Independent mathematicians are now translating the Lean code into human-readable mathematics. Buzzard expects that within a week or two, they have a much better understanding of what's actually going on in this code [6]. Andrew Sutherland of MIT, who independently compiled the Lean archive, called it "an historic moment" but emphasized: "There is much work to be done to understand the mathematical details of the proof" [1].

If the proof holds, it would be the second Millennium Prize problem solved in two months—both with heavy AI involvement—and would instantly validate hundreds of conditional theorems across number theory and cryptography [2][6]. If it fails, it joins a growing list of high-profile AI-assisted math claims that have not withstood scrutiny.

For now, the Riemann hypothesis remains officially unsolved. The evidence is a 17-million-line Lean archive that compiles; the verification is underway; the claim is unverified.

Sources

  1. Anthropic mathematician proves Riemann hypothesis, most famous ... (scientificamerican.com)
  2. Has the Riemann hypothesis been solved? Anthropic mathematician ... (clarin.com)
  3. Anthropic mathematician claims to have solved prestigious Millennium ... (theverge.com)
  4. Anthropic worker claims to have solved a Holy Grail of math problems with the help of AI | CNN Business (cnn.com)
  5. Anthropic worker claims to have solved famed 167-year-old maths problem with help of AI (yahoo.com)
  6. Riemann hypothesis: Has the world’s most famous maths problem been cracked? | Science and Technology News | Al Jazeera (aljazeera.com)
  7. AI Models Help Prove Navier-Stokes Blows Up in Finite Time (emergent.sh)
  8. Anthropic researcher uses AI to solve math's greatest riddle (dw.com)
  9. Anthropic worker claims to have solved a Holy Grail of math problems with the help of AI (whro.org)
  10. Anthropic worker claims to have solved a Holy Grail of math problems with the help of AI (kanw.org)
  11. Anthropic worker claims to have solved a Holy Grail of math problems with the help of AI | WAMC (wamc.org)
  12. Anthropic worker claims to have solved a Holy Grail of math problems with AI's help - ABC News (abcnews.com)
  13. https://alpo.ge/remark.pdf (alpo.ge)
  14. Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis \ Anthropic (anthropic.com)
  15. GitHub - leanprover-community/riemann-hypothesis: Independent Comparator setup for the Lean formalization of the Riemann Hypothesis (Alpöge, 2026-11-20) · GitHub (github.com)