Four months after an announcement that AI would have access to data stored in a ubiquitous cloud workspace solution, this happens…

In April 2026, Google had introduced Workspace Intelligence as the layer that gives its Gemini agent a real-time understanding of work content and removes the need for users to supply context every time they ask a question. Unless an administrator turns off the relevant sources in the Admin console, Gemini will have default access to organizational data across Gmail, Chat, Calendar, and Drive, including Docs, Sheets, and Slides, as announced.

The practical effect is broad. When a user submits a prompt, Gemini can ground its answer in indexed Workspace data such as email, files, calendar entries, and chats, so long as those sources remain enabled, and the user already has permission to view the material. Google says the system respects existing access controls, and it does not search beyond what the user can already see.

Even so, the default-on setup has renewed concern among enterprise IT and compliance teams, because a feature meant to improve convenience can also widen the surface area of internal data exposure inside an organization, as mentioned by one ZDNET report on 17 August 2026.

The official justification

Google’s position is that the data is not used to train its generative AI models, and not used for advertising. Also:

  • Prompts and responses stay within the organization’s domain
  • The system does not build a separate AI database from documents or email.
  • The issue is less about external data monetization and more about internal policy, governance, and whether an organization is comfortable letting an AI system scan sensitive material by default.
  • For regulated sectors such as healthcare, finance, or any environment bound by data privacy regulations, the default configuration may still require tighter limits on certain sources.
  • Administrators who want to restrict the data access will need to use the Workspace Intelligence Sources panel to disable access by source. These controls can be applied at the domain, organizational-unit, or group level, but there is no direct per-user toggle in the current interface.
  • Organisations that want a narrow deployment have to structure it through admin hierarchy rather than through a simple user-level switch.
  • Google recommends leaving all data sources enabled will preserve the best AI experience across Workspace, while the principle of least privilege in data privacy management dictates security teams to limit unnecessary access.

Now, four months after the announcement, various experts have crystalized their concerns. The tension is straightforward: Google is optimizing for AI usefulness, while enterprise administrators are being asked to decide how much internal data should be automatically available to a system that can summarize, retrieve, and reframe it on demand.