AI news · Anthropic · research
Claude cracks the 25-year riddle of NIST’s curve seeds
The numbers behind five standard cryptographic curves came from a stock random-number generator seeded with the clock in 1997 and 1998.
On 16 October 2026, Anthropic said an internal Claude research model had solved a 25-year cryptography puzzle in under half an hour. A great deal of online security rests on a few elliptic curves, mathematical objects behind digital signatures and secure web connections.
The US standards body NIST published five of them in 2000. Nobody could say where the starting numbers behind them, called seeds, came from. The NSA had supplied them, and that gap fed a long suspicion that the agency had chosen them to hide a weakness.
The answer is mundane. The seeds came from lrand48, an ordinary random-number generator that ships with the C programming language. Each run was started from the computer clock (a Unix timestamp) at a moment between May 1997 and April 1998. That retires a long-told story, recalled by a veteran NSA employee, that the seeds were a scrambled version of a joke phrase.
The finding has an outside check. Filippo Valsorda, a cryptographer who ran a public bounty for the answer, rebuilt the method himself. The Anthropic researchers collected the reward on 15 October and gave it to ISRG, the charity that runs Let’s Encrypt, which made it three times larger, at $36,864. Valsorda writes that the result rules out a known theoretical way to rig the seeds.
The signatures and secure connections built on these curves now rest on seeds anyone can trace to a clock.
Sources
- Anthropic: Claude finds the origins of the NIST ECDSA curve seeds. Checked
- Filippo Valsorda: The NIST Elliptic Curves Seeds Bounty Has Been Claimed. Checked
More AI news from
- India says lab AI agents read firms’ confidential files
- China drafts export curbs on 23 AI technologies
- Tim Cook says China has approved Apple Intelligence
- Reflection releases Beam’s weights under an open licence
- Campaigns spend $137 million on AI ads, most with no label
- Anthropic says juniors are a shrinking share of its tech hires
- Trump’s Super Intelligence Force orders a report on AI risks
- Microsoft finishes MAI-Base-2, a 2.8 trillion parameter model
- Apple tightens App Store rules on AI nudify apps