When creating a Microsoft Purview Communication Compliance policy using the "Detect Microsoft Copilot interactions" template, the system is pre-configured to monitor for specific risks associated with generative AI. This template automatically adds two specific trainable classifiers to the policy: "Unauthorized disclosure" and "Protected Materials". These classifiers are designed to detect instances where users might be sharing sensitive, confidential, or proprietary information with Microsoft Copilot, helping organizations enforce data security and compliance policies.

B. Prompt Shields: This is a safety system within Microsoft Copilot that detects and blocks harmful user prompts, not a trainable classifier used in communication compliance policies.

C. Threat: This classifier is automatically included in the "Detect threats and harassment" policy template, which targets different types of inappropriate content.

D. Corporate Sabotage: This classifier is part of the "Detect corporate sabotage" policy template, designed to identify communications that signal sabotage activities.

1. Microsoft Learn. (2024). Detect Microsoft Copilot for Microsoft 365 interactions with communication compliance. Under the "Policy template" section, it explicitly states: "When you select the Detect Microsoft Copilot for Microsoft 365 interactions policy template, the following two trainable classifiers are automatically added to the policy: Unauthorized disclosure... Protected Materials...".

2. Microsoft Learn. (2024). Learn about trainable classifiers. The "Pre-trained classifiers" section describes the purpose of various classifiers, including "Threat," "Corporate Sabotage," "Unauthorized disclosure," and "Protected Materials," confirming their distinct roles.

3. Microsoft Learn. (2024). Create and manage communication compliance policies. The "Policy templates" section details which classifiers are associated with other templates, such as "Threat" for the "Detect threats and harassment" template.