Anthropic says its Mythos model found vulnerabilities in cryptographic algorithms that secure the internet

Anthropic researchers reported that their Claude Mythos Preview model identified vulnerabilities in HAWK, a post-quantum cryptographic signature scheme previously vetted by experts for two years. The AI discovered these weaknesses within 60 hours at an API cost of approximately $100,000. This result demonstrates that large language models are becoming capable of accelerating the discovery of flaws in complex security protocols that underpin modern internet infrastructure.