Learn how Microsoft Entra Tenant Governance helps you discover and govern tenants across your organization to establish an AI-ready secure foundation
Updated Aug 10, 2026
Version 2.0Right now it looks Agent 365 driven by permission scope and session activity rather than the agent's actual decision behavior over time. In the reliability work I have done, configuration drift and reasoning drift turned out to need very different detection models, one is a state diff, the other needs a behavioral baseline. Is there a roadmap for baselining agent decision quality the same way tenant settings are baselined, or does risk stay scoped to permissions and activity for now?