Anthropic's Claude AI model has traced the origins of NIST ECDSA curve seeds to Unix timestamps from the 1990s, solving a decades-old mystery. The breakthrough sheds light on the cryptographic standards widely used today.
TL;DR
- Anthropic's Claude model traced NIST ECDSA curve seeds to Unix timestamps from the 1990s.
- The discovery solves a longstanding mystery about the origins of these cryptographic standards.
- The findings do not indicate any weakness in the ECDSA curves, according to Anthropic.
What happened
Anthropic's Frontier Red Team reported on October 16, 2026, that an internal Claude research model traced the origins of the NIST ECDSA curve seeds. The seeds were found to be outputs of the C library's lrand48 random-number generator seeded with Unix timestamps from the 1990s.
The discovery was made by asking Claude to identify the origins of the seeds without providing the humorous-passphrase tip that had been previously suggested. In under half an hour, Claude identified the origins of the NIST P-192, P-224, P-256, P-384, and P-521 seeds.
Claude approached the problem by hypothesizing a pseudorandom number generator that shipped with 1990s C libraries. It scanned the published seeds bit by bit, testing adjacent chunks as potential consecutive outputs of a 1990s-vintage generator. The model identified a pattern that pointed to the lrand48 generator.
Why it matters
The discovery solves a longstanding mystery about the origins of the NIST ECDSA curve seeds, which had been provided by the NSA without explanation. This breakthrough increases confidence in the security of the cryptographic standards widely used today.
The findings do not represent a cryptographic break, a new attack, or any weakness in the ECDSA curves, according to Anthropic. The seeds were produced by a common pseudorandom number generator initialized from the clock, a method anyone can now reproduce.
The discovery leaves some questions open, such as why certain candidate seeds were skipped and the origins of the base points for some curves. However, the team does not consider the selection of weakened curves plausible.
Key facts
- Anthropic's Claude model traced the origins of NIST ECDSA curve seeds to Unix timestamps from the 1990s.
- The seeds were found to be outputs of the C library's lrand48 random-number generator.
- Claude identified the origins of the NIST P-192, P-224, P-256, P-384, and P-521 seeds in under half an hour.
- The discovery does not indicate any weakness in the ECDSA curves, according to Anthropic.
- The findings do not represent a cryptographic break, a new attack, or any weakness in the ECDSA curves.
- The team observed that some potentially viable candidate seeds were skipped, but does not consider the selection of weakened curves plausible.
- The $36,864 bounty for finding the seeds will be donated to the Internet Security Research Group.
- The work is credited to John Wood, Henry de Valence, Ava Howell, Zack Voase, Lucy Sweet, Amir Omidi, Paul Crowley, Iain McGinniss, Nicholas Carlini, and Steve Weis.
Context
ECDSA is the most widely used elliptic curve cryptography standard. NIST adopted it in 2000 as part of the FIPS 186-2 Digital Signature Standard, which lists standard elliptic curves of varying sizes and security levels.
The NSA provided the seeds for these curves with no explanation or justification, leading to speculation and a bounty for whoever could find their origins. The leading account held that longtime NSA employee Jerry Solinas had generated the seeds as the hash of a humorous passphrase.
Anthropic's breakthrough sheds light on the origins of these cryptographic standards and increases confidence in their security. The discovery also highlights the capabilities of AI models like Claude in solving complex problems.
