Anthropic's Claude Mythos cracks HAWK-512 but fails to dent AES
LLMs won't break symmetric crypto
Anthropic's LLM Claude Mythos found a key-recovery attack on the post-quantum signature candidate HAWK-512, reducing its security from 128 bits to at most 108, but the attack isn't practical. It also improved an attack on 7-round AES-128, posing no threat to full AES. The author argues LLMs won't break established symmetric crypto due to the lack of mathematical structure, reliance on differential cryptanalysis, and decades of scrutiny. LLMs may instead help find errors in proofs and explore new attack classes.
Imagine a farmer who drags a tractor out into a patch of quicksand, and then buries it under cement. That’s what symmetric cipher design is like; it’s deliberately designed to come up with structures that are quick and easy to apply, but very messy and hard to untangle.