AI Agent Autonomously Hacked FreeBSD in Four Hours

An AI agent hacked into FreeBSD — one of the most secure operating systems in existence — in just four hours, without any human assistance. The finding emerges from research published this week.

The agent, running on Claude, exploited a kernel vulnerability to hijack kernel threads, write shellcode, and spawn a root shell entirely on its own. It represents the most advanced autonomous offensive cyber operation by an AI system documented to date.

Nvidia CEO Jensen Huang recently told Lex Fridman that he believes AGI has already been achieved. If AI agents can execute fully autonomous intrusions into hardened systems in hours, Huang may have a point.

For CIOs, the implication is stark: the threat landscape is shifting at a pace that human security operations centers were not designed to handle. AI-driven offensive security can scale attacks in ways that render traditional patch cycle practices insufficient. Automated defense and continuous kernel-level monitoring should now be on the priority list for anyone running critical infrastructure.