Meta's AI safety chief lost control of her own agent. An AI bot attacked 6 open-source repos for 10 days undetected. Your controls were built for humans. Agents aren't human. The gap between executive confidence and actual capability is the real risk. https://securityboulevard.com/2026/03/metas-ai-safety-chief-couldnt-stop-her-own-agent-what-makes-you-think-you-can-stop-yours/