Anthropic's LLM Claude Mythos discovered a key-recovery attack on HAWK, a post-quantum signature scheme candidate, reducing its security estimate, and an improved attack on 7-round AES-128. The negative result—no breakthrough against full-round AES—is arguably more significant, reinforcing that LLMs won't break symmetric crypto. The author advocates for formalizing LLM-assisted cryptanalysis and unifying cryptographic formalism.
Background
Anthropic released Claude Mythos, an LLM specialized for cryptanalysis, and published papers demonstrating its ability to discover new attacks on cryptographic primitives like HAWK and AES.
- Source
- Lobsters
- Published
- Aug 2, 2026 at 05:11 AM
- Score
- 7.0 / 10