AI-powered assistant that helps security teams detect, investigate, and respond to threats
Copilot Studio is designed as an enterprise-grade, secure online service and includes multiple controls to protect sensitive information, but risk management still depends on how it is configured and governed in the organization.
Key security and compliance characteristics:
- Enterprise compliance and certifications
Copilot Studio is an Online Service under Microsoft’s Online Services Terms and is compliant with or covered by major frameworks, including HIPAA coverage, HITRUST CSF, FedRAMP, SOC, ISO certifications, PCI DSS, CSA STAR, UK G-Cloud, OSPAR, K-ISMS, Singapore MTCS Level 3, and Spain ENS High. These are designed to support handling of sensitive and regulated data when used in accordance with organizational policies. - Data residency and encryption
Copilot Studio supports geographic data residency, allowing agents to operate within specific locations so that data is stored and processed in-region to meet local regulatory requirements.
Underlying Azure services provide:- Encryption of data at rest and in transit using industry-standard protocols.
- Data replication within the same geographic region for durability, while respecting residency requirements.
- Strong access controls such as MFA and RBAC to limit access to sensitive data.
- Access control and least-privilege behavior
Copilot Studio is secure by default and tailors responses based on the authenticated user and their permissions. When agents access internal sources (for example, SharePoint sites with confidential content), users only see information they are authorized to access. This helps prevent oversharing of sensitive content. - Sensitivity labels and data loss prevention
- Copilot Studio supports Microsoft Purview sensitivity labels for multiple knowledge sources (SharePoint, OneDrive for Business, SQL, Dataverse, etc.). Labels can classify and protect content (for example, Confidential, Highly Confidential) without blocking collaboration.
- For SharePoint knowledge sources, data loss prevention (DLP) endpoint filtering is supported to help prevent leakage of sensitive information.
- When agents use labeled content, responses and citations display the highest-priority sensitivity label from the data used, helping users recognize when they are dealing with sensitive material.
- Additional security features in Copilot Studio
Copilot Studio incorporates:- Bot authentication and authorization (for example, OAuth 2.0) so only authorized users can interact with agents.
- Data masking capabilities for sensitive information in interactions, especially for voice-enabled agents where variables can be marked as sensitive.
- Audit logs and monitoring to track agent activities and user interactions, supporting detection and investigation of potential issues.
- Secure integration with other Microsoft services via secure APIs and connectors.
- Responsible AI and safety controls
Every interaction undergoes privacy, security, and compliance checks aligned with Microsoft’s Responsible AI principles. Components include:- Grounding in trusted, permission-trimmed organizational data.
- Content safety filters to detect and mitigate harmful or inappropriate content.
- Guardrails to keep agents within defined scopes and reduce unintended actions.
- Encryption, RBAC, secure APIs, and continuous vulnerability scanning as cybersecurity measures.
- Purview integration for broader protection
With Microsoft Purview, organizations can:- Use sensitivity labels and DLP to classify and protect data used by Copilot Studio.
- Apply endpoint DLP policies that can restrict Copilot Studio agents (when the knowledge source is SharePoint and scoped to the Microsoft 365 Copilot location) from processing content with specific sensitivity labels.
What this means for highly sensitive documents:
- Copilot Studio and its underlying Azure services provide strong security, compliance, encryption, and access control capabilities suitable for sensitive and regulated data when configured correctly.
- Risk of information leakage is mitigated by:
- Permission-trimmed access to data sources.
- Sensitivity labels and DLP policies.
- Encryption and strict access controls.
- Monitoring and audit capabilities.
However, overall risk depends on:
- How data sources, labels, and DLP policies are configured in Microsoft Purview and Microsoft 365.
- How agents are designed (for example, which knowledge sources they can access and what actions they can perform).
- Organizational governance, including who can build agents, how they are tested, and how usage is monitored.
For highly sensitive information, it is recommended that security and compliance teams:
- Enforce appropriate sensitivity labels and DLP policies on all relevant content.
- Restrict agent access only to necessary data sources.
- Use audit logs and analytics to continuously monitor agent behavior and usage.
References:
- Ensure compliance with Copilot Studio
- Security and geographic data residency in Copilot Studio
- Security FAQs for Copilot Studio
- Use Microsoft Purview to manage data security & compliance for Microsoft Copilot Studio
- View sensitivity labels in agent responses (preview)
- Application Card: Microsoft Copilot Studio