When an AI Agent Causes an Incident, Can We Actually Investigate It? The Governance Gap Nobody Is Talking About Yet

Imagine an AI agent with access to your email, calendar, and purchasing systems takes an action you didn't authorise. Maybe it was prompt-injected. Maybe it misunderstood a task. Maybe something in it