Google is weaving its Gemini generative‑AI into Gmail, Drive and other Workspace apps, turning the assistant into a default feature rather than an optional add‑on. The model processes the content you feed it for “isolated tasks,” then discards the data, but the same outputs—summaries, snippets, or draft emails—can be recycled into training sets. Users who want absolute privacy must avoid any Gemini interaction altogether.

Google’s $185 billion AI investment through 2026 fuels its push to embed Gemini, though it says the model never uses Drive files or email bodies to train models. A “Gemini Apps Activity” switch, when disabled, stops storing chat history and thus blocks training, but also erases interactions, forcing users to choose between a chat record and chance their data will be fed back into the system.

The real friction comes from the UI: the privacy toggle lives in an obscure “Activity” submenu, not in Google’s main account settings, and disabling Gemini forces the loss of other Smart Features like inbox categorisation, Smart Compose and package tracking. This trade‑off illustrates a classic dark‑pattern—protecting user agency only if they endure a degraded experience—leaving many to wonder if the convenience of AI is worth the hidden cost.