Copilot Studio Compliance and Data Privacy | Day 88

Day 88 of 365 Days of Copilot Studio looks at compliance and data privacy across the life of an AI agent. An agent may retrieve business documents, receive personal information in conversations, and pass data to connected services. Each path needs an owner and appropriate controls.

Following Day 87: Security Best Practices, this lesson turns the source PDF into a practical privacy review for planning, deployment, and ongoing operation.

Start with responsibilities and data flows

Identify the business purpose, intended audience, and information the agent needs. Map where data comes from, where it is processed or stored, who can access it, and where actions send it. Include knowledge sources, connectors, channels, conversation records, and downstream systems.

Work with privacy, legal, security, and business owners to identify applicable data protection obligations and organizational policies. Requirements may differ by data, jurisdiction, and industry. Document decisions and evidence so they can be reviewed as the agent changes.

Classify before connecting

Use your organization’s classification scheme to distinguish public, internal, confidential, and restricted information. These are illustrative categories; use the labels and definitions your organization actually applies. Personal information, customer records, financial data, and intellectual property can require different handling.

Assign an owner to each source and confirm that its classification, audience, and permissions fit the agent’s purpose. Remove or isolate content that should not be available to the agent.

Minimize collection and sharing

Collect only the information necessary for the use case. Avoid asking for unnecessary personal data, scope retrieval to approved sources, and restrict transfers to approved destinations. Review the information that can appear in conversations and action outputs as well as the source documents.

Minimization reduces the amount of sensitive data that must be protected, retained, and reviewed; it does not replace access controls or a retention policy.

Protect knowledge and control data paths

Review permissions at each knowledge source and verify the authentication and retrieval behavior for its configuration. Test with representative authorized and unauthorized users in the actual channel. The SharePoint security guide provides related permission checks.

Review applicable Power Platform data policies for supported connectors, combinations, endpoints, channels, and agent capabilities. Confirm which controls apply to the actual integration rather than assuming a policy covers every path. The PDF’s examples of approved and blocked services illustrate organizational policy choices.

Use governance tools where supported

Microsoft Purview capabilities can help organizations classify information, prevent certain data loss, review audit activity, and manage retention. Coverage varies by workload, licensing, configuration, and supported integration. Identify the specific controls available for each relevant data location and validate their behavior before relying on them.

Review conversations and retention

Conversation records can contain user questions, responses, personal information, or business data. Determine what is stored in your environment and channels, who can access it, and what storage and retention settings apply. Protect transcripts as business data.

Define retention and deletion requirements before deployment with the responsible owners. Review approved storage locations, legal holds or other applicable requirements, and the available disposal process. Test and periodically revisit the configuration rather than assuming one setting covers every copy or connected service.

Verify residency and connected services

Check the geography of the relevant Microsoft 365 or Power Platform environment and the data residency commitments for each service. Identify where connected services store and process information, including whether a data flow can cross a regional boundary. Do not assume all services process data in the same region.

Audit, validate, and monitor

Review available audit events and operational records for relevant access, administrative activity, and changes to the agent, knowledge sources, or connections. Audit coverage depends on supported events, configuration, and licensing. Record gaps, establish ownership for review, and protect the logs themselves.

Before release, test permissions, data policies, sensitive data exposure, retention behavior, and relevant data paths. Keep evidence of the results and retest after significant changes to sources, channels, integrations, or business use.

Privacy readiness checklist

• Data and sources are inventoried and classified.
• Collection and sharing are limited to the stated purpose.
• Source and conversation access permissions are reviewed and tested.
• Applicable data policies and supported controls are validated.
• Conversation storage, retention, and disposal requirements are assessed.
• Residency and processing locations are checked for connected services.
• Available audit coverage, review responsibilities, and gaps are recorded.
• Privacy tests are completed and reviewed before release.

Privacy is an ongoing design responsibility. Revisit the checklist as the agent, its data, and its integrations evolve.