Beyond Users and Workloads: The Agentic Shift in Cloud-Native Zero Trust

Cloud-native zero trust matured in two waves. First, companies stopped trusting the network and moved the control plane to identity. Then, the same principle extended to workloads through managed identities, scoped role assignments and microsegmentation between services. Both waves worked on the assumption the actor is deterministic, so you can scope it, gate it and make sense of what it did afterwards.

The advent of agentic AI breaks that assumption. The platforms are catching up, and agent identities are now first-class principals that appear in sign-in logs and can be targeted by policy. Least privilege defeats its own purpose when the agent's entire value lies in deciding for itself what it needs. Meanwhile, the authorization decision has quietly migrated out of the policy engine and into the model.

This session traces the shift from securing users and workloads to securing autonomous cloud actors. Viewers will leave with an architectural path grounded in the Cloud Security Alliance Agentic Trust Framework.

Key Takeaways: - How to give agents real identity. - Move authorization back out of the model. - Carry provenance through delegation chains. - Close the gap between overseeing agents with ability to stop them.