Chrome quietly removed documentation claiming its on-device AI features do not send user data to Google servers.
Key Takeaways
Chrome’s on-device AI was previously marketed with an explicit privacy claim that data stays local; that claim has been deleted.
The removal implies Chrome AI features may transmit browsing or interaction data to Google infrastructure.
No public changelog or explanation accompanied the removal, leaving default data-sharing behavior undisclosed.
Users relying on the on-device framing for privacy decisions now have no official documentation to reference.
Hacker News Comment Review
Commenters broadly read this as deliberate: embedding AI in a browser with 3+ billion installs is a scalable, low-friction mechanism to harvest behavioral data from users who never read privacy policies.
The thread frames AI model hosting as a data pipeline business, with model quality as the enterprise product and raw user data as the real asset sold to advertisers or insurers.
Debate surfaced around Chromium-based alternatives: Brave’s Leo is cited as privacy-oriented but not on-device, and Brave itself carries historical reputation baggage that keeps many users on Firefox.
Notable Comments
@cferry: “the AI business is all about data collection… the amount of data that comes ‘for free’ to whoever hosts the models” – frames AI hosting as a data acquisition strategy, not a product.
@Animats: asks whether Google defaulted the data-sending feature to off – implying the answer is obviously no.
