Chrome Deletes Its Own Privacy Promise for Sneaky On-Device AI
In Chrome version 147, the settings page stated that AI features such as scam detection operated on-device “without sending your data to Google servers.” This language was removed in the newly released Chrome 148, now only stating that certain AI features run on-device, without mention of user data privacy. Users noted the change, sparking privacy debates online. Chrome has silently downloaded a large (roughly 4GB) Gemini Nano AI model file to eligible devices without prompting or notification; deleting the file leads to its automatic re-download. Privacy researcher Alexander Hanff documented this behavior on multiple operating systems, arguing it violates EU ePrivacy law requiring explicit user consent for storing data. The prior privacy claim served as Google’s justification for the silent installation, but Chrome’s AI Mode still routes some queries to Google’s servers, undermining the original promise. Removing the privacy wording does not resolve the legal or ethical concerns, but it ends Chrome’s explicit commitment to local-only data handling for these AI features. Google has not commented. The UI change is now effective for users on Chrome 148.
