Google removed privacy language from Chrome's on-device AI settings toggle in version 148, stating the change clarifies how APIs work rather than reflecting a change in data handling.

Chrome 148 removed a stipulation from the settings label stating that the on-device AI model would not send data to Google's servers, a change from version 147.

Google stated the wording change was made to clarify how AI works on the web: data passed to the local model is processed on-device, but if a website (including Google sites) uses the API, that website sees the input and output, and Google servers receive data only from Google websites.

The removal of the privacy assurance coincided with growing user backlash against AI features in 2026 and prompted users seeking to disable the 4GB AI model to question whether their data was being sent to Google's servers.