Microsoft Purview

Capabilities

• AI and Agent Activity Audit

  Observe Agents Users Data

  Captures audit logs of AI and agent activity in Microsoft Purview — user interactions plus agent authoring and configuration changes — so security teams can investigate what agents did and how they were changed. Reduces the risk of missing audit and undetected tampering.

• Communication Compliance for AI Interactions

  GovernObserve Users Data

  Detects inappropriate, risky, or policy-violating content in AI prompts and responses — including Microsoft 365 Copilot, connected AI apps, and browser-accessed third-party AI tools — using built-in classifiers and custom policies.

• eDiscovery for AI Interactions

  Govern Users Agents Data

  Lets legal and compliance teams search, place holds on, review, and export the prompts and responses of Microsoft 365 Copilot, agents, and other AI apps in Microsoft Purview eDiscovery — treating a user mailbox or an agent instance as the custodian. Reduces the risk of being unable to preserve or investigate AI interaction data for legal and compliance cases.

• Enterprise AI App Connectors

  ObserveGovern Users Data

  Connects sanctioned non-Microsoft enterprise AI apps — ChatGPT Enterprise and Anthropic Claude (connector in preview) — to Microsoft Purview so their prompts and responses surface in Data Security Posture Management for AI with auditing, and, for ChatGPT Enterprise, insider risk, communication compliance, eDiscovery, and retention. Brings enterprise AI outside Microsoft under monitoring and compliance, though sensitivity-label, encryption, and DLP enforcement are not supported for these connected apps.

• Entra-registered AI App Governance

  GovernSecure Data

  Registers custom, in-house AI apps in Microsoft Entra and integrates them with the Microsoft Purview SDK so their prompts and responses inherit the full Purview control plane — data classification, sensitivity labels, encryption, data loss prevention, insider risk, communication compliance, eDiscovery, retention, and Compliance Manager. Gives apps you build the same data governance as Microsoft Copilot rather than leaving them outside compliance.

• Sensitive Information Types and Classifiers

  ObserveSecure Data Users

  Identifies sensitive data with pattern-based sensitive information types (SITs) and machine-learning trainable classifiers. In AI scenarios these classifiers find sensitive content in user prompts and responses and provide the detection basis that labels, DLP, DSPM reports, and investigation all build on.