An Anthropic AI model took about half an hour to trace the origin of five seed values used to generate NIST elliptic curves. The NSA supplied the seeds in the late 1990s without explanation, fuelling years of suspicion that they might hide a backdoor. According to the finding, they came from an ordinary random number program fed with timestamps from 1997 and 1998. Users of the curves do not need to take action, and Bitcoin is unaffected because it uses a different curve.
What it means: The finding matters in two ways. First, it reduces a long-standing doubt about curves that protect much of the internet, from HTTPS connections to digital signatures. If the seeds came from a simple generator with timestamps, the theory that the NSA tried many options until it found a weak curve becomes less likely. Some parameters remain unexplained, though, so the debate is not fully closed. Second, the case shows AI models starting to contribute to real research, solving in minutes problems that kept specialists busy for years. That is good news for security, but the reverse also holds: the same tools can help find weaknesses in cryptographic implementations. Greek businesses do not need to do anything right now. The real task for the coming years is the move to quantum-resistant algorithms, for which the EU has already set a timeline. Organisations should map where they use elliptic curve cryptography so they are ready when the time to switch comes.
Related service: Custom software
Related stories
- Replit switches its default model to OpenAI, displacing Anthropic
- OpenAI model spent from someone else's test wallet to finish a task
- Microsoft exits Nscale's Essex AI data centre over power supply problems
- Meta removes terms language that seemed to allow AI training on business websites
- Drupal fixes two critical remote code execution flaws and urges immediate updates