When you create a communication compliance policy in Microsoft Purview and select "Detect Microsoft Copilot interactions," certain trainable classifiers are automatically added to help detect sensitive or inappropriate AI usage.
The "Unauthorized disclosure" classifier helps detect cases where users might share confidential or sensitive information via Copilot interactions, preventing unintended data leaks. The "Protected Materials" classifier is used to identify sensitive or restricted content that should not be shared through Copilot, ensuring compliance with organizational policies.