AI Hacking Agents Reach 69.3% in New Test, Exposing a Growing Security Automation Risk

An AI model that can write malicious code is dangerous only if a person or another system can deploy it effectively. An autonomous penetration ...