Building Agents You Can Trust on Windows | BRK262

Agents now run real commands, modify files, and move data. This breakout highlights the full set of Windows primitives that help agents discover, reason about, and execute AI‑powered activities across the system, while still operating within clear boundaries. Through live demos, you’ll see how agents identify available capabilities, plan and take meaningful actions, and how Windows layers permission scoping, inspection, developer tool capabilities, and rollback to keep developers in control.
Seating for this session is first-come, first-served. Add it to your schedule to plan your day and arrive early to secure a spot.

To learn more, please check out these resources:
* https://aka.ms/build26-next-steps
𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀:
* Kirupa Chinnathambi
* Stuart Schaefer
* Patrick Nikoletich
𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻:
This is one of many sessions from the Microsoft Build 2026 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com

BRK262 | English (US) | Windows
Breakout | (300) Advanced
#MSBuild

Chapters:
0:00 - Introduction and Session Overview
00:01:03 - Current Problems with AI Agents and Risks of Autonomy
00:03:04 - GitHub Copilot Insights and Permission Mismatch Issues
00:05:00 - Windows Role in Making Agents Safe by Design
00:06:03 - Agent Identity Concept and Demonstration
00:09:07 - Containment Principle and Microsoft Execution Containers (MXC) Introduction
00:14:14 - MXC Demonstration: Sandbox Configuration and Malicious Agent Test
00:19:03 - Agent Manageability and Continuous Supervision
00:21:00 - GitHub Copilot Sandboxing Integration and CLI Demos
00:26:12 - Summary: Identity, Containment, Manageability and Windows Ecosystem Integration

Видео Building Agents You Can Trust on Windows | BRK262 канала Microsoft Developer