Anthropic announced on October 16, 2026 that an internal Claude research model solved, in under half an hour, a question that had circulated among cryptographers for about 25 years: the origin of the “seeds” used to generate the elliptic curves standardized by NIST in the ECDSA standard.
What happened?
The P-192, P-224, P-256, P-384 and P-521 curves, widely used in digital signatures and a large part of internet infrastructure, were adopted in 2000 based on seeds supplied by the NSA without explanation. This fueled speculation about possible backdoors. The best-known rumor, attributed to the late NSA mathematician Jerry Solinas, was that the seeds were the hash of a humorous phrase. Several Anthropic researchers had tried to follow that lead without success.
In October 2026, a company employee asked an internal research model how the seeds were generated, without mentioning the rumor. Claude identified that they did not come from any hash, but from the output of the lrand48 random number generator in the 1990s C library, initialized with Unix timestamps between May 1997 and April 1998. For the P-256 curve, for example, the seed corresponds to candidate 1,842 of a run started on August 16, 1997 at 01:52:34 UTC.
Why it matters
According to Anthropic itself, the result is not a cryptographic break, a new attack or a weakness in the curves. On the contrary, the simplicity of the method (a common PRNG fed by the system clock) makes the idea of a deliberate selection of weak parameters less plausible. The company published Python code that allows the seeds to be reproduced from the timestamps and candidate index.
Cryptographer Filippo Valsorda, who had organized a bounty for the mystery, independently confirmed the result and reported that the Anthropic team claimed the prize on October 15. The $36,864 was donated to the Internet Security Research Group (ISRG), the organization behind Let's Encrypt.
What has not yet been confirmed
Not all details have been clarified. The seed of the binary curve B-163 remains unexplained, as do the base points of some important curves (including P-256). Anthropic notes that some viable candidates were skipped during generation, but considers it unlikely that this indicates the selection of weakened curves.
Sources
- Official Anthropic post: Claude reveals ECDSA curve origins
- Filippo Valsorda's confirmation of the bounty
Transparency: This content was created, edited or reviewed with the aid of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.
By GeekikiBot