Microsoft Purview Adds Plaintext AI Prompt Monitoring for IT Admins

Microsoft announced a new security capability for its Purview platform on August 17, 2025. The update introduces plaintext monitoring of artificial intelligence prompts and responses within the Insider Risk Management module. This feature allows authorized personnel to view potentially risky messages sent to AI services. Messages containing specific risk indicators will be surfaced directly to IT teams in readable text format.

System employs pseudonymization by default while security members can deanonymize users with authorization

The system employs a pseudonymization mechanism by default to protect user identity. Security members can only deanonymize a user if they believe an AI interaction violated company policies. This deanonymization process requires specific authorization from the administrator. Role-based access controls and audit logs remain in effect to prevent misuse of this powerful visibility tool.

Microsoft plans to release a preview of this feature later in August 2025. General availability is scheduled for September 2025. The rollout applies to global regions. The company referenced this update in Message Center document MC1304292 to inform enterprise administrators of the upcoming changes.

This development addresses the growing need for enterprise governance over shadow AI usage. Organizations must balance data leakage prevention and intellectual property protection with employee privacy. The ability to inspect plaintext interactions helps IT teams identify compliance risks that were previously hidden behind anonymized logs. This move signals a shift toward stricter internal controls on generative AI adoption in the workplace.

Microsoft has not confirmed the specific launch window for the preview release beyond the current month. The vendor has not disclosed pricing details for this specific feature set. Information regarding regional restrictions for the preview phase remains unconfirmed.