Adam Back Says NSA Should Have Published NIST Curve Generation Steps
An internal Claude model traced five NIST curve seeds to a random-number generator initialized with Unix timestamps. The analysis identified no cryptographic break, new attack or weakness.

Blockstream CEO Adam Back said the National Security Agency (NSA) should have published reproducible steps for generating five elliptic curves standardized by the National Institute of Standards and Technology (NIST), after a model reconstructed how their seeds were created.
“That's NSA fault for not revealing through reproducible steps to generate their NIST published curves,” Back wrote on Oct. 17.
On Oct. 16, an internal Claude model traced the seeds for NIST P-192, P-224, P-256, P-384 and P-521 to lrand48, a C-library random-number generator initialized with Unix timestamps. The model identified the method in less than half an hour. For P-256, candidate 1,842 matched Aug. 15, 1997, at 9:52:34 p.m. ET (Aug. 16, 1997, at 01:52:34 UTC).
The reconstruction also mapped five other prime-curve seeds from the ANSI X9.62 standard to Unix timestamps between May 1997 and April 1998. NIST’s 2000 Digital Signature Standard listed the curve parameters. Under its procedure, a seed’s hash is used to derive a curve coefficient; the reconstruction traced how the seed itself was generated.
Ledger co-founder Nicolas Bacca joked about the time spent choosing a more complex curve, writing: “Yes but we lost so much time to end up on a more complex curve … which was the joke.”
The findings concern NIST curves, not Bitcoin’s curve. Bitcoin (BTC) software uses secp256k1, which is separate from the five NIST curves examined. The reconstruction does not explain why Satoshi Nakamoto chose secp256k1. It identified no cryptographic break, new attack or weakness, and users have no action to take.
Some questions remain open. The analysis found that some potentially viable candidate seeds were skipped but did not establish why. It also did not explain the origin of the B-163 binary-curve seed or how base points were selected for P-224, P-256, P-384 and P-521. The pattern makes deliberate weakening less plausible, but does not resolve every question about the curves’ selection.