AI agents are being handed the keys to legacy estates, sensitive citizen data and shared services, often before anyone has decided who's accountable for them. They're curious, tireless and eager to act, but they don't always know their limits, and they rarely ask permission. Under tight budgets, stretched teams and intense audit scrutiny, that's a risk few can afford. This session explores why identity is the control plane for agentic AI: proving who (or what) each agent is, what it can touch, and for how long. Securing agentic AI isn't child's play, but with identity in control, it can be.