Anthropic's AI Finds New Crypto Weaknesses, Throws Post-Quantum Transition into Question

Anthropic's AI model, Claude Mythos Preview, has discovered new weaknesses in HAWK and AES cryptographic algorithms. The findings do not affect today's production systems, as HAWK is still a research proposal and the AES attack only applies to a reduced-round version used for academic research.

Claude found a previously unknown mathematical symmetry in HAWK after 60 hours of research, reducing the estimated work needed to recover the smallest HAWK key from 2^64 operations to 2^38. This is roughly 67 million times less computation than previously believed.

The practical fix for HAWK would be to double its key size, but this change would remove many of the efficiency advantages that made HAWK an attractive post-quantum candidate.