Building Agents You Can Trust on Windows | BRK262

This Microsoft Build 2026 session demonstrates how Windows provides primitives for building trustworthy AI agents that can safely discover capabilities, reason about actions, and execute tasks within controlled boundaries. Through live demos, developers learn about agent identity, containment via Microsoft Execution Containers (MXC), permission scoping, and rollback capabilities. The session is designed for advanced developers building AI agents on Windows who need robust security and control mechanisms.