Anthropic says Claude found new weaknesses in HAWK and reduced-round AES

ccerrato147 · x · 2026-07-29

Anthropic says Claude Mythos Preview helped researchers find two cryptographic weaknesses: one that significantly weakens HAWK, a post-quantum signature scheme, and another that attacks round-reduced AES.

The company stresses that no deployed systems are affected, there is no break of AES itself, and no standardized post-quantum cryptography has been broken. The main implication is that frontier models can now uncover mathematical flaws in cryptographic schemes, not just implementation bugs.

Related event: Anthropic's Claude Mythos Uncovers Flaws in Post-Quantum Crypto and AES(30 posts)→

Original post →

More from Safety

Safety channel →