Employees paste customer data, source code, and internal documents into AI tools every day. What happens to that content afterward often comes down to a detail nobody at the organization is watching: whether the person was signed in. Google Gemini does not use prompts or responses to train its models while a user is signed in, but a signed-out session carries no such assurance.
DefensX now closes that gap. Under AI Protections, Disable Data Sharing for Model Improvements can be enabled for Google Gemini. Once enabled, users must log in to Gemini before they can use it, so every prompt is submitted under an authenticated session and stays out of model training.
This feature requires a CORE+ subscription or higher and is available in extension version v4.2.222 / v4.3.222 or later.