- Published on
‘Wow, it’s a real discovery’: Claude cracks a 25-year NSA number mystery
- Authors
- Name
- DeepHorus
An Anthropic AI model has cracked a 25-year-old cryptography mystery, tracing five unexplained NIST elliptic-curve seeds to an ordinary random-number generator. The seeds were supplied by the National Security Agency in the late 1990s and were long suspected of hiding a backdoor. The model, Claude, reportedly solved the puzzle in under 30 minutes, making a hidden NSA backdoor less likely.
The numbers underpin elliptic-curve cryptography, which protects much of the internet and bitcoin wallets. In 2000, NIST adopted curves provided by the NSA without explaining how some starting values, called seeds, were chosen. After the Snowden disclosures linked the NSA to weakening cryptographic standards, suspicion deepened and NIST later removed the Dual EC random-number generator from its recommendations.
Cryptographers had sought an answer for years. Neal Koblitz and Alfred Menezes examined the controversy in a 2015 paper, "A Riddle Wrapped in an Enigma." Security researcher Steve Weis filed public-records requests in 2018, but NIST said it had no documentation and the NSA stopped responding.
In 2023, cryptography engineer Filippo Valsorda and other contributors offered a $12,288 bounty, rising to over $36,000 if donated to charity. NSA mathematician Jerry Solinas, who sent the seeds in 1997, died earlier that year. The puzzle remained unsolved.
This month, an Anthropic employee asked an internal research model to investigate without sharing the old phrase theory. Claude traced the seeds to lrand48, a random-number generator in older C libraries, using timestamps from 1997 and 1998. It reproduced the sequence behind P-256 and found an exact match at the 1,842nd candidate. "Wow, it's a real discovery," one observer said.
The result does not require users of NIST curves to take action, though questions remain about how some elements were selected. Bitcoin was never affected because it uses secp256k1, a simpler curve with little room to hide a weakness. Still, sufficiently powerful quantum computers could eventually pose a separate threat. The discovery may close one chapter in cryptographic history while highlighting how AI can accelerate mathematical detective work.
