Editorial illustration for Anthropic's Claude cracks 25-year-old NSA curve mystery in under 30 minutes
Claude Solves 25-Year NSA Cryptography Mystery in 30 Min
Anthropic's Claude cracks 25-year-old NSA curve mystery in under 30 minutes
Anthropic's Claude needed less than 30 minutes to solve a cryptography puzzle that cryptographers have poked at for 25 years. The question dates back to 2000, when NIST standardized several elliptic curves as part of ECDSA, now the most common framework for elliptic-curve digital signatures. The NSA handed over the seed values used to generate those curves but never explained where the numbers came from.
That silence bred suspicion. If the NSA had picked the seeds to secretly favor curves with exploitable weaknesses, the signatures built on top of them could be compromised without anyone knowing. The mystery got attention well beyond academic circles, drawing a cash bounty and years of amateur attempts to reverse-engineer the seeds.
Jerry Solinas, an NSA employee who worked on the original standard, once suggested the seeds were derived from a hash of a joke phrase. He died before confirming it, and later analysis showed he was wrong anyway.
That left the origin of the seeds unresolved for a quarter century, right up until Anthropic decided to point its AI model at the problem.
The seeds came from "lrand48," a simple C random number generator, initialized with a Unix timestamp between May 1997 and April 1998. Claude spotted the pattern because "lrand48" only outputs 31 bits, and the top bits in some seeds were systematically zero. According to Anthropic, this makes deliberate backdoors less likely.
Why this matters
For twenty-five years, a question about the NSA's elliptic curves sat unresolved because the person who could answer it, Jerry Solinas, had died, and the rest of the cryptography community treated it as a problem requiring specialized human archaeology. Claude cleared it in under 30 minutes. That gap, decades of amateur sleuthing and a standing bounty versus half an hour of machine reasoning, is the real headline for anyone building with these models.
If an LLM can reconstruct a cryptographer's undocumented methodology from indirect evidence, the same capability applies to auditing your own codebase, your vendor's crypto implementation, or that legacy system nobody fully understands anymore. For founders, this is a preview of what "due diligence" looks like when AI can retroactively explain decisions nobody wrote down. For researchers, it's a reminder that provenance and documentation matter more, not less, once your work can be reverse-engineered on demand.
We'd be cautious about overreading one well-publicized case, but the pattern is worth tracking: mysteries that outlasted human patience are now solvable in a coffee break.
Common Questions Answered
What cryptography mystery did Anthropic's Claude solve and how long did it take?
Claude solved a 25-year-old NSA elliptic curve mystery in under 30 minutes, determining the origin of seed values that the NSA had used to generate curves standardized by NIST in 2000 as part of ECDSA. The cryptography community had been unable to solve this puzzle for decades, with the question remaining unresolved even after Jerry Solinas, the person who could have answered it, passed away.
What was the source of the NSA's elliptic curve seeds according to Claude's findings?
Claude discovered that the seeds came from 'lrand48,' a simple C random number generator that was initialized with a Unix timestamp between May 1997 and April 1998. The AI spotted the pattern by identifying that 'lrand48' only outputs 31 bits, and the top bits in some seeds were systematically zero, which Anthropic suggests makes deliberate backdoors less likely.
Why had the NSA's elliptic curve seed selection remained a mystery for 25 years?
The NSA handed over the seed values used to generate the standardized elliptic curves but never explained where the numbers came from, which bred suspicion in the cryptography community about whether the NSA had deliberately chosen seeds to secretly favor curves with hidden weaknesses. The mystery persisted because specialized knowledge and human archaeology were thought to be required to solve it, and the person most likely to have the answer, Jerry Solinas, had died.
What does Claude's rapid solution of this cryptography puzzle reveal about AI capabilities?
The fact that Claude solved in 30 minutes what cryptographers couldn't solve in 25 years demonstrates a significant gap between machine reasoning and traditional human expertise in pattern recognition and specialized problem-solving. This capability highlights that AI models can potentially reconstruct complex technical knowledge and solve problems that were previously treated as requiring deep domain expertise and historical investigation.
Further Reading
- Anthropic’s Claude cracks a 25-year NSA curve mystery, what it means for bitcoin - CoinDesk
- Anthropic says Claude solved 25-year cryptography mystery in 30 minutes - Yahoo Finance
- Claude AI Cracks the NSA’s 25-Year Cryptography Seed Mystery - Bitcoin.com News
- Anthropic's Claude Solves The NSA Elliptic Curve Mystery In 30 Minutes - Bitget
- Anthropic’s Claude AI Uncovers Origins of NSA-Provided Cryptographic Keys After 25 Years - GBHackers