Agentic AI is breaking identity as we know it. Autonomous systems now request access, assume roles, and mint credentials in real time—creating ephemeral, self-escalating privilege chains that traditional IAM cannot see or control.

Static roles, long-lived credentials, and human-centric governance models collapse in an environment where identities act continuously, negotiate access dynamically, and operate at machine speed. The result: an expanding, largely ungoverned identity attack surface.

This session reframes identity as a runtime control plane for agentic systems. We’ll examine patterns for enforcing continuous authentication, just-in-time authorization, and verifiable delegation across human and non-human identities. We’ll also explore how dynamic secrets, policy-driven access, and end-to-end auditability can contain privilege sprawl and restore control.
Walk away with a concrete architecture for extending Zero Trust to AI-driven workloads—where every token is ephemeral, every action is policy-bound, and every decision is attributable.