Google lets Gemini access Workspace data by default, so turn it off

what Google does behind the scenes

Google Workspace gives Gemini automatic access to an organization’s Gmail, Docs, Calendar, Chat, Drive and Meet from the moment the service is enabled. The company states that the information is not used to train the model, does not leave the organization’s domain, and is not compiled into a dedicated AI knowledge base. In practice, this means any document, email or calendar event can be retrieved in real time the instant an employee asks the chatbot a question.

how it works technically

The mechanism is based on Retrieval-Augmented Generation. Google indexes the organization’s information sources—what it calls Workspace Intelligence Sources—just as it has always done for internal search. When a request arrives, the system performs a real-time search over the materials the user is authorized to see, pulls the relevant excerpts, and feeds them to the model for answer generation. It is not training, it is not separate storage; it is search with a generation layer on top.

why it worries organizations

The risk is not an external leak but uncontrolled internal exposure. Customer contracts or regulations may prohibit AI systems from scanning information even within the organization. Moreover, departmental isolation can break down: an answer generated for a sales employee could include details from a human-resources complaint or a confidential deal intended only for senior management. Google does not block this by default; the responsibility falls to administrators.

how to turn it off

Workspace administrators can disable the intelligence sources per service or globally through the admin console. The setting resides under Apps > Google Workspace > Gemini Settings, where access to each service can be revoked individually. Once disabled, Gemini stops pulling information from that source, but the existing index remains for normal search.