AI Traces NIST Curve Seeds to 1990s RNG Code in 30 Minutes
Anthropic’s Claude model identified the source of five standard cryptographic seeds, easing fears of hidden NSA backdoors in digital signature algorithms.
Key points
- Claude determined NIST curve seeds originated from the lrand48 random number generator.
- The findings were published in a report dated October 16, 2026, by Anthropic researchers.
- Anthropic is donating the associated $36,864 bounty to the Internet Security Research Group.
Anthropic reported that its Claude AI model solved a 25-year-old cryptography puzzle regarding the origin of NIST elliptic curve seeds. The model completed the analysis in under 30 minutes, identifying the specific random number generator used to create these standard parameters.
The seeds for curves P-192, P-224, P-256, P-384, and P-521 were traced to lrand48, a C library function initialized with Unix timestamps from 1997 to 1998. This finding addresses long-standing suspicions that the NSA might have embedded weaknesses into the Elliptic Curve Digital Signature Algorithm standard adopted in 2000.
Historical Context
The mystery began when the NSA provided seeds for digital signature curves without explaining their creation method. Former NSA employee Jerry Solinas had suggested the values were hashes of humorous phrases, a theory that guided earlier research. Claude’s analysis disproved this by matching the seeds to output patterns typical of 1990s software development tools.
Affected Standards and Services
The five curves identified are part of the widely used ECDSA framework defined in NIST FIPS 186-2. These parameters underpin trust in digital signatures that verify data integrity and identity. Cyber Security News noted that services like Apple’s PQ3 iMessage protocol use P-256, highlighting the broad operational reliance on these specific mathematical constants.
Defender's checklist
Organizations should review their cryptographic implementations to confirm they are using validated libraries rather than custom code. Security teams should monitor vendor announcements for any changes to signature verification processes. Auditors should verify that current systems align with updated standards regarding curve selection and parameter generation methods.
Operational impact
Leaders should ask whether their suppliers rely on legacy cryptographic libraries that may require re-validation. Teams usually review certificate chains and signing keys to ensure they remain compliant with updated trust models. Managers need to brief legal and compliance staff on how the origin of these parameters affects long-term data integrity policies and vendor risk assessments.
What to do and how to stay safe: Anthropic
- Audit your software stack to ensure cryptographic libraries are up to date and support standard curves.
- Verify that digital signature verification processes use trusted, well-maintained implementations.
- Monitor vendor advisories for any recommendations regarding curve parameters or signature algorithms.
- Confirm that your organization's data retention policies account for the lifespan of current cryptographic standards.
Step-by-step guide: Incident Response Plans: Why Your Security Tools Fail Without One
General security guidance from the InfoSec Ledger newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Did this discovery break existing encryption?
No, the finding explains how public parameters were chosen and does not expose secret keys or weaken current encryption.
Which random number generator was identified?
The seeds were traced to lrand48, a simple C random number generator initialized with Unix timestamps.
What happened to the bounty?
Anthropic is donating the $36,864 bounty to the Internet Security Research Group.
