Claude cracks NIST curve seed mystery in 30 min
Original: Claude reveals ECDSA curve origins
Why This Matters
Resolves a foundational trust question about one of the world's most-used crypto standards.
An Anthropic internal research model traced the origins of NIST P-192 through P-521 elliptic curve seeds to Unix-timestamped lrand48 RNG output—ending decades of cryptographic speculation in under 30 minutes.
For over two decades, cryptographers have wondered where the NSA-supplied seed values for NIST's ECDSA elliptic curves came from. The seeds—used to define curve coefficients in the FIPS 186-2 standard adopted in 2000—were submitted with zero explanation, fueling persistent suspicion about potential backdoors. NSA employee Jerry Solinas suggested in 2015 the seeds came from hashing a humorous internal phrase like 'Bob and Jerry rule,' but he died in 2023 without confirming details, and no archives were found.
In October 2026, an Anthropic employee asked an internal Claude research model to investigate the seed origins without providing any prior clues. In under 30 minutes, Claude determined that none of the seeds involved any hash function at all. Instead, they are the direct output of C's lrand48 pseudo-random number generator, seeded with a Unix timestamp. The algorithm sampled 20-byte candidates sequentially—P-256 required 1,842 iterations before landing on its standardized seed value.
Anthropic is clear this is not a cryptographic break. The finding actually strengthens confidence in the curves: a simple RNG initialized from a system clock is exactly the kind of mundane, uncontrolled process that makes deliberate backdooring implausible. 'Most researchers trusted the curves already,' the post notes. No action or configuration changes are needed.