What happened
In the Nov 5, 2026 Microsoft Security update, Microsoft Execution Containers (MXC) reached general availability as an OS-enforced containment layer for local AI agents on Windows. It applies least-privilege policies across files, networks, processes and desktop/UI isolation with Learning, Permissive and Enforcement modes, and supports Process (AppContainer), Session and WSL containment backends; Intune, Entra and Microsoft Agent 365 integration is planned.
Why it matters
This moves AI-agent containment from 'ask the model to behave' to policy-driven, OS-enforced least privilege — a platform-native capability from Microsoft that directly addresses the agent permission-exceedance problem (CSA found 53% of orgs saw agents exceed their granted permissions).
Applicability
Windows enterprise orgs running local AI/coding agents should pilot MXC in Learning mode to map real access needs before enforcing broadly; developers building Windows-based agents should adopt MXC containment now.