Contoso Healthcare has deployed AI agents in Microsoft 365 that access sensitive patient data stored in SharePoint. The security team wants to ensure these agents can only access resources when running from trusted, compliant execution environments. Which control should the administrator implement to enforce this requirement?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Conditional Access isn't just for people — it works for agents too. Point a CA policy at your agent identities and you can require that they only operate from trusted, compliant locations. It's like putting a bouncer on your data that checks agent credentials at the door.
Full explanation below image
Full Explanation
Microsoft Entra Conditional Access policies can target agent identities (service principals and managed identities registered through Entra Agent ID) just as they can target human user accounts. By creating a Conditional Access policy scoped to the agent service principals, administrators can enforce conditions such as trusted IP locations (e.g., only from approved Azure regions or on-premises environments), compliant platforms, and additional authentication controls. This ensures agents can only execute and access resources from approved environments, significantly reducing the attack surface if an agent identity is compromised.
Option A, sensitivity labels on SharePoint, control who can access and use labeled content based on user identity and label policy. However, sensitivity labels do not evaluate environmental compliance or location of an agent's execution — they check label permissions, not network location or device compliance.
Option C, a SharePoint permission group with manual approval, provides basic access control but is not scalable, does not enforce execution environment compliance, and cannot dynamically enforce conditions like network location or platform compliance in real time.
Option D, Microsoft Purview Insider Risk Management, monitors for risky user behavior patterns using signals from Microsoft 365 activity. It is a detection and investigation tool, not a preventive access control mechanism for agent identities.
Exam tip: Conditional Access is the right answer whenever a question asks about enforcing environmental or contextual access conditions (location, device compliance, network) for identities — including agent identities. Remember that CA policies in Entra can target non-human identities (agents, service principals) as well as users.