LLMs won't break symmetric crypto
Summary
The article discusses Anthropic's cryptanalytic work using Claude Mythos and argues that LLMs are unlikely to break established symmetric cryptography. It breaks down two notable attacks on HAWK and AES-128, notes that the AES attack is not practical, and highlights the robustness of symmetric schemes due to their high-level structure and differential cryptanalysis. The piece references CryptanalysisBench and Matthew Green's analysis, concluding that LLMs will not derail current cryptographic guarantees.