A proof of the Riemann hypothesis would not break RSA, elliptic-curve cryptography or Bitcoin keys. Anthropic mathematician Levent Alpöge said on 20 November 2026 that he and the AI model Claude had found one, and released it as computer-checkable Lean code that other mathematicians are still verifying. Encryption rests on different hard problems, and a proof gives an attacker no shortcut for any of them.
In brief
- The proof is unverified. Terence Tao describes it as an enormous proof that exists only as Lean code and expects it to take some time to digest.
- RSA depends on large numbers being hard to factor, and elliptic-curve cryptography on a problem called the discrete logarithm. Neither assumes the Riemann hypothesis.
- The software that generates encryption keys uses primality tests whose guarantees need no unproven hypothesis.
- If the hypothesis is true, every algorithm that relies on it already works today. A proof adds certainty for mathematicians and no new capability for attackers.
- The scheduled threat to RSA and elliptic curves is the quantum computer. NIST plans to deprecate both after 2030 and disallow them after 2035.
What has been claimed about the Riemann hypothesis?
Levent Alpöge, a mathematician at Anthropic, says he believes he has proved the Riemann hypothesis and its generalised form while working with the company’s Claude model. The note on his website dates the work from 6 October to 16 November 2026 and credits Claude with much of the technical progress, saying the model “far outpaced me near the finish line”. Instead of a conventional paper, the proof is a Lean archive split into fourteen files, and the note refers to an informal written version of more than 250,000 pages.
The Riemann hypothesis dates from 1859 and concerns the zeros of the zeta function, which control how evenly prime numbers are spread among the integers. It is one of the seven Millennium Prize Problems, each carrying a $1 million prize from the Clay Mathematics Institute. The generalised version extends the same statement to a wider family of functions.
Has the Riemann hypothesis proof been verified?
Nobody had confirmed the proof as of 21:00 UTC on 20 November 2026. Terence Tao wrote on his blog that Alpöge had generated an enormous, Lean-only proof of the Generalized Riemann Hypothesis using an internal AI model. He pointed readers to a thread in the Lean community’s chat as the place to follow the checking, and said he imagined that digesting the proof would take some time.
Lean is a proof assistant, a program that checks every logical step of an argument written in its language. Two things still need confirming: that the code compiles, and that the theorem it ends on is the Riemann hypothesis and not something weaker. On the second, Elliot Glazer said the formal statement is the Generalized Riemann Hypothesis and that the code was still compiling, according to OfficeChai’s round-up of reactions. Scientific American reported that early, informal responses from mathematicians were positive.
A prize would take years, because the Clay Mathematics Institute’s rules require a solution to be published in a qualifying outlet, to win general acceptance among mathematicians, and to have been in print for at least two years.
Does a proof of the Riemann hypothesis break RSA encryption?
A proof of the Riemann hypothesis does not break RSA, because RSA’s security rests on how hard it is to split a very large number into its two prime factors, and a proof supplies no faster way to do that. The best known factoring algorithms for ordinary computers do not assume the hypothesis, according to an explainer Safeguard published in October, so confirming it changes nothing about how they run.
Algorithms that do lean on the hypothesis are no more dangerous after a proof. If the hypothesis is true, they give correct answers today, whether or not anyone has written down why, and an attacker could have used them last year. A proof changes what mathematicians know and leaves what a computer can do unchanged.
New mathematics sometimes leads to new algorithms years later, and a proof this large will contain new techniques. Nobody has shown a route from this work to faster factoring, and Alpöge’s note does not mention cryptography.
Is Bitcoin or elliptic-curve cryptography at risk?
Bitcoin and elliptic-curve cryptography are no more exposed to a proof than RSA is. Elliptic-curve keys, which are widely used to secure web traffic and to sign Bitcoin transactions, depend on the difficulty of the discrete logarithm problem on a curve. Bitcoin mining depends on the SHA-256 hash function. Neither rests on the Riemann hypothesis.
Free tool
How does your own site score?
Run the same forty passive checks against your own domain: TLS and certificates, security headers, SPF and DMARC, cookies before consent, and what your stack quietly reveals. A grade out of 100 in about fifteen seconds.
No signup. Nothing installed. We only request what your site already serves publicly.
What in cryptography depends on the Riemann hypothesis?
Very little that is in everyday use depends on the Riemann hypothesis. The main example is a primality test from 1976, and key-generation software uses a different one.
| Area | Depends on the Riemann hypothesis | What a proof changes |
|---|---|---|
| RSA keys | No. Security rests on factoring | Nothing known |
| Elliptic-curve keys, including Bitcoin signatures | No. Security rests on discrete logarithms | Nothing known |
| Hash functions such as SHA-256 | No | Nothing |
| Generating primes for keys | No. The NIST standard FIPS 186-5 requires the randomised Miller-Rabin test, whose error bounds are proved without any hypothesis | Nothing |
| Miller’s deterministic primality test (1976) | Yes. Its correctness assumes the generalised hypothesis | The test becomes unconditional. An unconditional test, AKS, was published in 2004 |
Why do people think a proof would break encryption?
The belief comes from the fact that the hypothesis and RSA both involve prime numbers, and from two decades of headlines that joined the two. In 2004 a newspaper report on an earlier claimed proof said that all codes could become breakable and no internet transaction would be safe. A cryptographer on a mailing list that week questioned the report, writing that he knew of no factoring or discrete-logarithm algorithm that would become drastically faster. When Sir Michael Atiyah announced a proof in 2018 that mathematicians did not accept, press coverage again said a proof could crack the foundations of modern cryptography.
What does threaten encryption?
Two things threaten encrypted systems in practice: quantum computers in the future, and stolen logins now. Shor’s algorithm, published in 1995, would let a large enough quantum computer factor numbers and solve discrete logarithms quickly, which breaks both RSA and elliptic curves. NIST published its first post-quantum standards in August 2024, and its transition plan, finalised on 27 October 2026, deprecates quantum-vulnerable public-key algorithms after 2030 and disallows them after 2035.
Attackers do not need to break encryption when they can sign in. The November snapshot of the Ransomnews infostealer dataset found employee credentials in stealer logs for 1,663 of 1,856 large organisations, 344,017 credentials in all. A stolen session cookie gets past multi-factor authentication without touching the encryption underneath.
Two reports this week show where AI is changing attacks. The FBI warned that some intruders using the AI penetration testing tool ARTEX have reached privileged access in under three hours, and Dragos found that secrets planted in AI prompts were stolen from three breached providers.
What should security teams do?
Nothing needs patching or replacing because of this claim, whether or not the proof holds.
- Leave RSA and elliptic-curve deployments as they are. No key length or certificate is weaker today than it was yesterday.
- Keep the post-quantum migration on the timetable NIST has set, with 2030 and 2035 as the dates that matter.
- When someone says the proof breaks encryption, ask which step of which algorithm it speeds up. Safeguard’s explainer suggests the same test.
- Check your domain for employee logins in stealer logs with StealerCheck. A stolen login gets an attacker in without breaking any encryption.
Frequently asked questions
Has the Riemann hypothesis been solved?
A proof has been claimed and not yet confirmed. Levent Alpöge of Anthropic released a Lean formalisation on 20 November 2026 that he says proves the Riemann hypothesis and its generalised form. Mathematicians are checking it.
Does a Riemann hypothesis proof break encryption?
No. RSA depends on factoring being hard and elliptic-curve cryptography on discrete logarithms. A proof of the Riemann hypothesis gives no faster method for either problem.
Is Bitcoin safe if the Riemann hypothesis is proved?
Yes. Bitcoin signatures use elliptic-curve cryptography and mining uses the SHA-256 hash function. Neither depends on the Riemann hypothesis.
What is Lean, and does a Lean proof mean the result is correct?
Lean is a proof assistant that checks each logical step of a proof written in its language. A Lean proof that compiles is very strong evidence, but people still have to confirm that the statement it proves is the one claimed.
What did Claude do in the proof?
Alpöge’s note credits Claude with much of the technical mathematics and says the model outpaced him near the end. Terence Tao describes the proof as generated with an internal AI model. Neither says which model version was used.
Who gets the $1 million Millennium Prize?
Nobody yet. The Clay Mathematics Institute requires a solution to be published in a qualifying outlet, to be generally accepted by mathematicians, and to have been in print for at least two years before it considers a prize.
Sources and further reading
- Levent Alpöge, note on the claimed proof, November 2026
- Terence Tao, Unforced Euler singularity, unforced Navier-Stokes singularity, and the Generalized Riemann Hypothesis, 20 November 2026
- Scientific American, Anthropic mathematician claims to prove Riemann hypothesis, most famous problem in all of math, 20 November 2026
- OfficeChai, How mathematicians have reacted to the claimed Riemann hypothesis proof, 20 November 2026
- Safeguard, The Riemann hypothesis explained: primes, zeta and cryptography, 23 October 2026
- NIST, IR 8547: Transition to Post-Quantum Cryptography Standards, October 2026
- Clay Mathematics Institute, Rules for the Millennium Prize Problems
- Ransomnews, How stealer logs power modern ransomware attacks
Reporting by the Ransomnews Research Team, from Alpöge’s note, Terence Tao’s blog and the other sources listed above, read on the evening of 20 November 2026. Ransomnews has not checked the proof and takes no position on whether it is correct. Stealer-log figures are aggregate counts from the 1 November 2026 snapshot of the Ransomnews Infostealer Exposure Dataset, which is built on the Alerts.bar index.
The Ransomnews Monthly
One email a month. Original leak-site data, victim census updates, and the findings that did not make the articles. No spam, unsubscribe any time.
Double opt-in. We store your email, signup time, and IP for consent records (GDPR Art. 7). See our privacy policy.
