Background and source of the controversy (Nov 23, 2025)
ZDNet's Nov 23, 2025 report drew attention by suggesting Google might read Gmail messages to train its AI models. In response the same day, Google clarified it did not use ordinary consumers' Gmail message bodies as a data source for general AI models (such as Gemini).
Where did the confusion arise?
ZDNet said the confusion mainly stemmed from how Google described data‑use policies across different products. Google’s documents described three categories of data sources: publicly available data, licensed/permissioned datasets, and enterprise data that in specific circumstances and with explicit consent from businesses or administrators may be used to improve models. For ordinary users, the latter two are not equivalent to personal Gmail; however, the initial wording did not clearly distinguish “Workspace (enterprise) data may be used for customized models with admin consent” from “personal Gmail is not used to train general models,” causing readers to misunderstand.
Google’s core clarification
Google told ZDNet that the data used to train general AI models mainly comes from publicly available web content, licensed data, and specific partner data; consumer Gmail content is not used to train those general models. For Google Workspace (enterprise) customers, administrators can choose whether to allow enterprise data to be used for model improvement or internal customization — this is an enterprise‑level opt‑in, not a default.
Historical note: ads and scanning policy
ZDNet also reminded readers that Google stopped automatically scanning Gmail message bodies for ad purposes as early as 2017. That policy change is a different issue from current debates about AI training: one concerns ad‑targeting data practices, the other concerns sources and compliance of data used for model training.
What ordinary users can do: opt‑out paths
If you still worry your data is being used for AI features, ZDNet lists several places to check and settings to turn off:
- Sign in to your Google account and go to the "Data & privacy" or "Personalization" settings to review AI‑related toggles.
- In Gmail settings, look for options like "Smart features & personalization" and disable permissions that allow using mail content to enhance personalized experiences or experimental AI features.
- Enterprise users should contact their Workspace administrator; admins can view and disable options in the Google Admin console that allow enterprise data to be used for model training or AI customization.
Conclusion and impact
ZDNet’s report and Google’s response together show that differences in wording during information dissemination can quickly amplify public privacy concerns. For younger users, the focus is not only on "what the company said" but also on how to manage data‑use rights in personal settings. The Nov 23, 2025 episode reminds companies they must be clearer when publishing data‑use policies and reminds users to regularly check account privacy and personalization settings.
For step‑by‑step instructions, ZDNet’s article end provides more detailed setup paths and screenshots; privacy‑conscious readers are advised to follow those prompts item by item.