SATURDAY · 21 NOV 2026 · 10:47 UTCRSSMY BRIEFINGS
RESEARCH21 NOV 2026 · 10:02 UTC · 7 MIN READ

Anthropic Mathematician Claims Riemann Hypothesis Proof With Claude

Anthropic mathematician Levent Alpöge claims he proved the Riemann Hypothesis using the company's Claude model, posting the unverified proof on his personal site November 20, 2026. Anthropic has not endorsed the claim, which remains subject to years of peer review by the Clay Mathematics Institute before any $1 million prize could be awarded. The proof, formalized in Lean, also purports to extend to the Generalized Riemann Hypothesis.

RESEARCH DESK
Anthropic Mathematician Claims Riemann Hypothesis Proof With Claude
Anthropic mathematician Levent Alpöge claims he proved the Riemann Hypothesis using the company's Claude model, posting the unveri…
Anthropic Mathematician Claims Riemann Hypothesis Proof With Claude

Short answer: Anthropic mathematician Levent Alpöge claims he proved the Riemann Hypothesis using the company's Claude model, posting the unverified proof on his personal site November 20, 2026. Anthropic has not endorsed the claim, which remains subject to years of peer review by the Clay Mathematics Institute before any $1 million prize could be awarded. The proof, formalized in Lean, also purports to extend to the Generalized Riemann Hypothesis.

Anthropic mathematician proves Riemann hypothesis with Claude

Anthropic mathematician Levent Alpöge says he’s cracked the Riemann hypothesis, one of math’s most stubborn unsolved problems, with a heavy assist from the company’s Claude model. The claim landed on Alpöge’s personal site on November 20, 2026; it hasn’t been independently verified, and Anthropic itself hasn’t commented or endorsed the announcement. If the proof holds, it marks a landmark moment for AI in mathematical research, arriving just months after OpenAI made a similar splash about the Navier-Stokes equations.

The Riemann hypothesis has resisted solution since Bernhard Riemann proposed it back in 1859. It deals with the distribution of prime numbers and the behavior of zeros in the zeta function, and it sits among the seven Millennium Prize problems curated by the Clay Mathematics Institute. Each carries a one-million-dollar reward, but the institute’s rules demand a proposed solution be published and then survive at least two years of scrutiny while gaining general acceptance across the global math community. In the twenty-five years since the prizes were established, only the Poincaré conjecture has been officially resolved.

Alpöge’s document states that Claude was integral to developing the argument and that the method extends to the Generalized Riemann Hypothesis, with the Extended Riemann Hypothesis also within reach. He writes that he and Claude have formalized the reasoning without assumptions using Lean, a proof assistant language. The mathematician also notes he texted Anthropic chief executive Dario Amodei about the result late on a Sunday night and suggests he delayed publication until after the company’s initial public offering on Thursday, November 19. Anthropic told *The Verge* in August it was exploring the Riemann hypothesis with Claude but didn’t expect the techniques in use to yield a complete proof.

Mathematicians struggle to verify AI-generated proof

The mathematical community now faces the daunting task of checking a proof that, by Alpöge’s own description, covers vast new territory. The Clay Institute’s two-year clock hasn’t started because the work hasn’t been accepted for publication in a recognized venue. Mathematicians will need to translate the Lean formalization into human-readable mathematics, verify each step, and assess whether the approach genuinely resolves the hypothesis or contains subtle gaps. That process could take years, and the history of the problem is littered with claimed proofs that collapsed under examination.

This episode follows a turbulent period for AI-driven mathematics. In September, OpenAI announced it had cracked the Navier-Stokes problem, another Millennium Prize challenge, using ten thousand agents and a multi-million-dollar compute budget. That claim has been met with cautious credibility but also with allegations that the system incorporated unpublished work from mathematicians who had shared research with OpenAI’s tools. The company has denied wrongdoing, but the controversy has deepened mistrust. Alpöge himself was collaborating with New York University professor Tristan Buckmaster on Navier-Stokes before OpenAI’s announcement, and Buckmaster has said the AI-produced solution followed the same route they were pursuing.

AI-assisted math claims overwhelm verification efforts

Alpöge also claimed in July that he had solved the Jacobian conjecture using Claude, adding to a deluge of AI-assisted results that has left many mathematicians struggling to keep pace. The sheer volume of material, combined with the opacity of how large language models arrive at their suggestions, is forcing a reckoning over attribution, verification, and the very nature of mathematical discovery. Researchers who build or deploy AI systems for formal reasoning now confront a landscape where breakthroughs may arrive faster than the community can absorb them, and where the line between human insight and machine-generated conjecture is increasingly blurred.

Engineering takeaways from AI mathematics breakthrough claims

For engineers and product teams working with AI, the immediate takeaway is caution. A claim posted on a personal website, even from a credentialed researcher at a leading lab, does not constitute a verified result. The formalization in Lean is a positive signal-it means the argument can be mechanically checked-but mechanical checking is only as reliable as the translation of the mathematical concepts into code. Teams should treat the announcement as a research artifact worth monitoring, not as a settled fact. They should also watch how the community handles the verification process, because the tools and norms that emerge will shape how future AI-generated proofs are evaluated, published, and trusted.

More broadly, the episode underscores that AI is becoming a co-author in the highest-stakes domains of knowledge work. The Riemann hypothesis sits at the center of number theory, with implications for cryptography, complexity theory, and the distribution of primes that underlie much of modern security. A genuine proof would reverberate far beyond pure mathematics. But until the Clay Institute or a comparable body certifies the result, the only responsible posture is skeptical patience. Builders should invest in tooling that supports rigorous formal verification, establish clear provenance chains for AI-assisted derivations, and prepare for a future in which the bottleneck isn’t generating candidate proofs but validating them.

Frequently asked questions

Has the Riemann hypothesis been officially proven by Anthropic or Levent Alpöge?

No. Levent Alpöge posted a claimed proof on his personal website on November 20, 2026, stating he used Anthropic's Claude model to develop it. However, the proof has not been independently verified, Anthropic has not endorsed the claim, and the Clay Mathematics Institute has not accepted it for publication or started its two-year verification period.

What is the Clay Mathematics Institute's process for validating a Millennium Prize solution?

The Clay Institute requires a proposed solution to be published in a recognized venue and then survive at least two years of scrutiny while gaining general acceptance across the global mathematics community. Only the Poincaré conjecture has met this standard in the 25 years since the prizes were established.

How was AI involved in Alpöge's claimed proof of the Riemann hypothesis?

Alpöge states that Anthropic's Claude model was integral to developing the argument. He and Claude formalized the reasoning without assumptions using Lean, a proof assistant language. The method reportedly extends to the Generalized Riemann Hypothesis, with the Extended Riemann Hypothesis also within reach.

Why are mathematicians skeptical of recent AI-assisted proofs like this one?

The mathematical community has a history of claimed proofs collapsing under examination. Recent controversies include OpenAI's Navier-Stokes claim, which faced allegations of incorporating unpublished work from mathematicians who shared research with OpenAI's tools. The volume of AI-generated material and opacity of how models arrive at suggestions has created verification challenges.

What should engineers and product teams take away from this announcement?

Treat the claim as a research artifact worth monitoring, not a verified result. While Lean formalization allows mechanical checking, this is only as reliable as the translation of mathematical concepts into code. Teams should invest in rigorous formal verification tooling, establish clear provenance chains for AI-assisted derivations, and prepare for a future where validating proofs is the bottleneck, not generating them.

Source: The Verge

Signal Blocks Gemini and Siri From Reading Screens
policy

Signal Blocks Gemini and Siri From Reading Screens

Signal has disabled in-app screenshots on Android and iOS 27 to prevent OS-level AI agents like Google's Gemini and Apple's Siri from scanning private conversations. The November 20, 2026 change blocks these assistants from reading on-screen content. Users can re-enable screenshots in settings, but Signal warns this exposes messages to AI agents. President Meredith Whittaker says developers lack tools to stop AI agents from bypassing app-level privacy controls without degrading usabili

Meta Reconsiders 2030 Net Zero Goal After Record AI Emissions
industry

Meta Reconsiders 2030 Net Zero Goal After Record AI Emissions

Meta is re-evaluating its 2030 net-zero target after emissions surged nearly 40% to 21.7 million metric tonnes in 2025, driven by rapid AI infrastructure expansion. The company's net emissions rose 45% to 11.8 million tonnes, the steepest increase among Big Tech peers. Microsoft, Google, and Amazon reported smaller increases but have maintained their original sustainability timelines.

Hebbia Reports Gains in Financial Research With GPT-6 Astra
models

Hebbia Reports Gains in Financial Research With GPT-6 Astra

Hebbia tested a prerelease GPT-6 Astra in August-September 2026 and reports it improved financial research workflows. The model scored 13.6 percentage points higher than the best alternative on citing required source types, with 99.5 percent of cited excerpts relevant. It also met 97.9 percent of presentation requirements, a 14.4-point lead, ranking first on both visual polish and brief adherence. Hebbia says this could shrink analysts' overnight deck revisions to minutes of review.

Share this post
Report