ANTHROPIC: AI DISCOVERS NEW CRYPTOGRAPHY ATTACKS, WEAKENS POST-QUANTUM CANDIDATE
Anthropic says its Claude Mythos Preview AI discovered two new cryptographic attacks, including one that significantly weakens the HAWK post-quantum digital signature candidate and another that speeds up attacks on a reduced-round version of AES by 200-800x.
The company stressed neither finding affects production systems or breaks the full AES standard.
Instead, the research demonstrates AI’s growing ability to uncover mathematical flaws in cryptographic algorithms that have withstood years of expert review.
Anthropic says Claude developed the HAWK attack in roughly 60 hours and estimates each breakthrough cost about $100,000 in API compute.

















