Discovering cryptographic weaknesses with Claude

This research shows that large language models can provide meaningful contributions in the field of cryptanalysis. The CryptanalysisBench metric, developed through a collaboration between ETH Zurich, Tel Aviv University and Haifa University, evaluates the ability of artificial intelligence models to systematically detect weaknesses in encryption algorithms. In the study, Claude uncovered previously unreported vulnerabilities in certain algorithms, indicating that AI can increasingly play a crucial role in security research.