2 Minute Drill: Building an Agentic AI Hacker with Drex DeFord

Drex unpacks a signal hiding across several recent stories: we may be deploying AI well ahead of our ability to secure it. From an autonomous AI agent that breached a consulting firm's internal chatbot in two hours, gaining access to 46 million employee messages, to a multi-agent experiment where AI systems escalated privileges, forged credentials, and disabled antivirus software entirely on their own, the pattern is unsettling. Add a publicly available AI prescribing tool in Utah that researchers convinced to triple an opioid dose and reveal its own system prompts, and the picture gets clearer. Prompts are the new attack surface and the security model hasn't caught up yet.