Anthropic says Claude solved NIST ECDSA curve seed mystery

By Splitfeed EditorialLast updated Oct 17, 2026, 05:00 UTC

On October 16, 2026, Anthropic announced that its Claude model had traced the origins of the seed values used in NIST's ECDSA elliptic curve standards, completing the analysis in under 30 minutes. The seeds were linked to 1990s timestamps and a random number generator from the same era. The finding addresses a cryptographic question that had gone unresolved for roughly 25 years, with the specific RNG identified in the process though its exact name has not been confirmed in the available reporting.

The stakes here are substantial because the NIST ECDSA curves underpin a large portion of the world's digital security infrastructure, from TLS connections to software signing and cryptocurrency protocols. For decades, cryptographers have questioned whether the seed values were chosen with a hidden purpose, a concern sharpened by the 2013 revelations around the Dual EC DRBG algorithm, which involved NIST standards and alleged NSA influence. If Claude's finding holds up to independent scrutiny, it would not definitively rule out backdoors in the curves themselves, but it would offer the first concrete, verifiable account of where the seeds came from, which is a meaningful step toward settling one of the more persistent suspicions in applied cryptography. What remains genuinely in dispute is whether the result has been independently verified and what exactly the identified RNG reveals about the intent behind the seed selection.

None of the three camps have published reactions to this story yet. The Pro-AI camp would typically treat a result like this as a showcase of AI solving hard, high-value problems that eluded human researchers for a generation, pointing to the speed and breadth of Claude's analysis as evidence of transformative capability. The Anti-AI camp would be expected to urge caution about accepting an AI-generated cryptographic finding without rigorous peer review, arguing that errors in this domain carry serious consequences. The Middle Ground camp would likely welcome the result as a promising research tool while stressing that the finding needs independent verification before it influences any security recommendations or policy.

The question to watch is whether academic cryptographers or NIST itself respond to Anthropic's claim with independent analysis. A peer-reviewed confirmation or refutation of the specific RNG identification would either validate Claude's result or reveal its limits, and that response will shape how seriously the security community incorporates AI-assisted cryptographic research going forward.