Here’s what Anthropic found when it turned Mythos loose on encryption algorithms

Anthropic's Claude Mythos AI identified mathematical weaknesses in HAWK, a post-quantum cryptographic candidate, and a simplified AES variant, representing a significant advancement in AI-assisted cryptographic research. While these vulnerabilities do not affect currently deployed systems and require impractical conditions to exploit, they highlight that even cutting-edge cryptographic standards can harbor subtle flaws that may become exploitable as technology advances. This matters because post-quantum cryptography is being actively standardized to protect future communications, and weaknesses discovered now are far less costly to address than those found after widespread deployment. Organizations must monitor cryptographic research closely and maintain agility in their cryptographic implementations to respond quickly when theoretical risks evolve into practical threats.