Claude traced NIST ECDSA curve seeds to a 1990s random number generator

By Splitfeed EditorialLast updated Oct 16, 2026, 20:40 UTC

On 16 October 2026, Anthropic announced that its Claude model had traced the origin of the seed values used in NIST's ECDSA elliptic curves, a question that had gone unanswered for roughly 25 years. According to Anthropic's account, Claude completed the investigation in approximately 30 minutes, identifying the seeds as deriving from a 1990s random number generator. The ECDSA curves were adopted by NIST without any public documentation of how their seeds were chosen, a gap that has persisted since the 1990s.

The significance of this finding reaches well beyond a puzzle being solved. NIST's ECDSA curves underpin a large portion of modern cryptographic infrastructure, including digital signatures used in software, financial systems, and secure communications. The undocumented seed origins have long been a source of suspicion in cryptographic research, with some researchers raising the possibility that the seeds could have been chosen in ways that introduced weaknesses, whether deliberately or not. Tracing them to a specific 1990s random number generator would, if independently verified, either substantiate or substantially reduce those concerns, depending on what that generator's properties turn out to be. What remains in genuine dispute is whether Claude's identification is correct, how it arrived at the answer, and what the cryptographic implications of that particular RNG actually are.

With no published reactions yet from any camp, it is possible to sketch what each would typically bring to a story like this. The Pro-AI camp would likely treat the 30-minute timeline as a striking demonstration of AI capability on a hard technical problem that defeated human researchers for decades. The Anti-AI camp would probably urge caution about accepting Anthropic's characterization at face value, pressing for independent peer review before any conclusions about cryptographic safety are drawn. The Middle Ground camp would likely argue that the result is genuinely interesting but that the real work begins now, with human cryptographers needed to validate the finding and assess whether the identified RNG introduced any exploitable bias into the curves.

The clearest next step is independent verification by cryptographers outside Anthropic. If researchers can confirm the RNG identification and analyze its statistical properties in the context of curve generation, that would either settle or sharpen the long-running debate about whether the NIST ECDSA curves carry hidden weaknesses.