Zero Trust architectures were designed around human users and static services. As AI agents begin executing actions across systems using delegated authority, these models are being stretched in new and sometimes unintended ways.

In this session, we will share practical insights from securing agentic AI systems operating in production environments. We will examine how identity, authentication, and authorization must evolve when agents act asynchronously, invoke tools, and operate across multiple trust boundaries. The session will highlight where traditional Zero Trust patterns still apply, where they need adaptation, and how security teams can preserve core Zero Trust principles without over-constraining agent-driven workflows.

This talk will focus on architectural considerations for maintaining trust boundaries and control in environments increasingly shaped by autonomous and semi-autonomous AI systems.