When AI Becomes the Attacker: LLM Agents Execute First Fully Autonomous Cyberattack

This attack marks a watershed moment in cybersecurity: for the first time, an LLM agent orchestrated an entire attack autonomously, adapting to environments and pivoting through infrastructure without human guidance. The speed and sophistication—full database exfiltration in under 60 minutes with zero schema knowledge—demonstrates that AI-powered threats have matured from theoretical risks to operational reality. Organizations must now prepare defenses against attackers that can learn, adapt, and exploit vulnerabilities faster than human operators.