Claude traces NIST ECDSA curve seeds to 1990s random number generator

By Splitfeed EditorialLast updated Oct 17, 2026, 13:00 UTC

On October 16, 2026, Anthropic reported that its Claude model had traced the origins of five NIST ECDSA elliptic curve seeds to timestamps and a random number generator from the 1990s, completing the analysis in under 30 minutes. The NIST ECDSA curves have been a source of suspicion in cryptography circles since their introduction in the 1990s, because the process used to generate their seeds was never made fully transparent, and the curves have historically been associated with NSA involvement in the NIST standardisation process. The finding was reported across technology and finance outlets beginning on October 16 and into October 17, 2026.

The significance of this finding rests on how long the question had gone unanswered. For roughly 25 years, cryptographers could not definitively account for how those specific curve seeds were chosen, which left open the theoretical possibility that they had been selected in a way that could introduce a backdoor or other vulnerability. Tracing them to timestamps and a concrete random number generator would, if verified, close that gap and provide a mundane rather than malicious explanation. What remains genuinely in dispute is the reliability and verifiability of Claude's analysis, since a result produced by an AI model in 30 minutes on a problem that resisted human cryptographers for decades invites scrutiny of the method, not just the conclusion. Reporting also notes that Bitcoin uses different elliptic curves and is not directly affected, a clarification that has been emphasised in coverage reaching cryptocurrency audiences.

The Pro-AI camp has treated this as a straightforward demonstration of AI capability, with Earth Insider on Bluesky framing it as Claude solving a 25-year mystery in minutes, emphasising speed and impact. The Anti-AI camp has not yet published reactions to this story, though it would typically raise questions about whether an AI-generated cryptographic analysis can be trusted without independent human verification, and whether Anthropic's framing overstates what was actually resolved. The Middle Ground camp has also not reacted publicly, but would be expected to welcome the result cautiously, asking for peer review of the method before treating the question as definitively closed.

The argument that would most directly settle the dispute is independent verification by cryptographers examining Claude's reasoning and methodology in detail. If the chain of evidence from the 1990s timestamps and random number generator to the five curve seeds holds up under expert scrutiny, the finding would carry considerably more weight than an Anthropic announcement alone.