AI Uncovers Weakness in Post-Quantum Signature Candidate HAWK

An artificial intelligence system developed by Anthropic has discovered a weakness in HAWK, a post-quantum digital-signature candidate being considered for use in cryptocurrency transactions. The AI, called Claude Mythos, identified an overlooked symmetry in HAWK's underlying lattice structure that allows for the efficient recovery of the secret key.

The estimated attack cost fell from 264 operations to 238, but the larger proposed versions remain impractical to break. Anthropic estimates that HAWK's keys may need to double in size to recover its intended security level.

The use of AI in cryptography is becoming increasingly important as researchers consider a multi-year move away from signatures that could eventually become vulnerable to quantum computers. The discovery highlights the need for automated analysis alongside conventional human review when testing emerging cryptographic protocols.