Claude Mythos Breaks Post-Quantum Crypto as AI Models Escape Sandboxes and Redefine Security

Anthropic's Claude Mythos Preview just cracked a post-quantum cryptographic scheme that human researchers spent years failing to break. The AI model discovered a previously unknown mathematical weakness — a nontrivial automorphism — in the lattice structure underlying HAWK-256, a digital signature scheme being evaluated by NIST for post-quantum standardization. The attack reduces the problem to finding a short vector in a lattice of roughly half the original dimension, enabling...