Anthropic says its Claude model has solved a cryptographic puzzle that has nagged researchers for decades: where the "seed" values behind NIST's elliptic curve standards actually came from. According to Anthropic, the seeds were not the product of a hash of a secret passphrase, and there is no sign they were crafted to hide an NSA backdoor. Instead, they were generated by lrand48, a pseudorandom number generator shipped with 1990s C libraries, initialized with a Unix timestamp.
The finding, published by Anthropic, covers the NIST P-192, P-224, P-256, P-384 and P-521 curves. The company says an internal research model cracked the question in under half an hour after a curious employee asked it to figure out how the seeds were made, without pointing it at the old rumor about a humorous passphrase. The result, Anthropic says, should be read as a fun closing chapter to years of speculation, not as a warning about the standard.
What Claude found about the NIST seeds
ECDSA is the most widely used elliptic curve cryptography standard, adopted by NIST in 2000 as part of the FIPS 186-2 Digital Signature Standard. That standard lists a set of curves at different sizes and security levels. For curves over prime and binary fields, the specifications publish a "seed" parameter whose hash value defines the coefficient b of the curve.
The problem was that the NSA supplied those seeds with no explanation. Researchers had no way to confirm they were chosen innocently rather than engineered to contain a deliberate weakness. The NIST P-256 seed, for example, was a string of hex characters with no stated origin. That gap fueled years of suspicion, including the influential paper "A Riddle Wrapped in an Enigma" by Neal Koblitz and Alfred Menezes, and even a bounty for anyone who could trace the seeds.
Anthropic says its employees had already chased the seeds after hearing they were a hash of a humorous phrase. That lead turned out to be wrong.



