For decades, experts questioned whether unexplained numbers in widely used security standards were concealed by a US National Security Agency (NSA) backdoor. Anthropic says its Claude AI traced them to a routine 1990s computer program and timestamps, without uncovering a security flaw. An Anthropic AI model has cracked one of cryptography’s longest-running mysteries, tracing numbers the NSA handed to the US government in the late 1990s - which were long suspected of hiding a backdoor - to an or
