Microsoft Execution Containers: Policy-driven containment for AI agents

Agents are unlocking enormous productivity gains for customers, but their ability to work across files, networks, and applications can introduce new security risks.