Back to all news
Developer ToolsJul 31, 2026

AI Models Stress-Test Cryptography Security update

Anthropic researchers used Claude to autonomously discover improved attacks on HAWK and reduced AES, highlighting AI’s potential to stress-test cryptographic algorithms.

Why now

This development underscores the need for accelerated cryptanalysis to address emerging AI-driven threats before deployment.

Key signals

Anthropic’s Claude autonomously discovered new attacks on HAWK and reduced AES.
AI models can synthesize existing cryptanalytic tools into new attacks.

Sources

Related coverage