J. Ayo Akinyele, head of engineering at Ripple, has acknowledged a major cryptographic breakthrough achieved by Anthropic’s Claude, an advanced artificial intelligence system developed by the AI research company Anthropic. The discovery addresses a decades-old mystery about the origins of popular cryptographic parameter standards.
AI identifies origins of NIST elliptic curve seeds
Anthropic revealed that an internal research version of Claude managed to uncover the origins of five widely used NIST elliptic curve seeds, including P-192, P-224, P-256, P-384, and P-521. These values are central to the Elliptic Curve Digital Signature Algorithm (ECDSA), a cryptographic standard employed in numerous applications, including securing cryptocurrencies.
For over 25 years, cryptography researchers questioned whether these seeds, provided by the US National Security Agency (NSA), might conceal hidden vulnerabilities. The lack of detailed documentation around the process intensified speculation and concerns over possible backdoors.
Claude AI found that the NIST elliptic curve seeds were generated via lrand48, a basic pseudorandom number generator present in older C programming libraries, using the Unix timestamp as a seed value. This finding provides a replicable explanation for the parameters and helps put to rest some longstanding uncertainty in the cryptographic community.
Mini dictionary: lrand48 – A pseudorandom number generator included in some early C language standard libraries. It produces a sequence of numbers based on a seed, commonly the system’s unix time, but is regarded as relatively basic by modern cryptographic standards.
AI research company Anthropic reported that Claude identified lrand48, seeded with a Unix timestamp, as the method used to generate five NIST elliptic curve seeds, providing reproducible evidence and greater confidence in the origins of these cryptographic standards.
Security impact and expert perspectives
Akinyele described the discovery as “remarkable,” noting that the actual explanation for the seeds’ generation is comparatively simple and suggesting it improves trust in the parameter selection. He also reflected on how artificial intelligence might shed light on other cryptography puzzles in the future.
He highlighted that the research results do not represent a vulnerability or an attack against the ECDSA standard, which underpins the security of many cryptocurrency systems. Instead, the findings make any theory that the NIST curves were intentionally weakened appear far less likely.
The discovery neither exposes ECDSA to attack nor introduces security risks for cryptocurrencies using elliptic curve cryptography, but it does enhance transparency and eases concerns over potential hidden flaws in these standards.
Anthropic pointed out that some uncertainties persist, particularly regarding why specific base points were selected and why some mathematically valid candidate curves were omitted from the final standards. While these aspects require further study, the central issue of the seed generation process has been clarified.
AI reshapes cryptographic research
The achievement demonstrates artificial intelligence’s growing role as a research tool in cryptography and cybersecurity. Experts believe such technology could accelerate discoveries, challenge old assumptions, and contribute to the transparency and trustworthiness of security systems employed worldwide, including in the digital asset sector.
The development is considered a step forward for those seeking to validate and improve the security foundations underlying the global crypto and blockchain ecosystem.
