Claude Mythos AI Model Uncovers Vulnerabilities in Post-Quantum Cryptography
A new study from Anthropic has shown that its Claude Mythos AI model can be used to strengthen security in the crypto sector by discovering vulnerabilities in post-quantum cryptography.
The researchers used the Claude Mythos Preview model to discover two cryptanalytic attacks - one on the post-quantum HAWK digital signature scheme and one on a reduced-round AES-128 variant. The results show that the model is capable of conducting cryptography research at an expert level, but they don't affect production systems.
The study found that Claude Mythos Preview discovered a previously unused symmetry in HAWK's mathematical structure that accelerated key recovery attacks. For HAWK-256, the expected cost of full key recovery dropped from 2^64 to 2^38 operations, while for HAWK-512 and HAWK-1024, the attack remains practically infeasible.
The researchers emphasized that the results don't require changes to deployed software and are part of a responsible disclosure process. The study also introduced the CryptanalysisBench benchmark to evaluate language models' cryptanalytic capabilities, which showed that five tested frontier models solved 65% to 86% of level-one tasks.