HeadlinesBriefing favicon HeadlinesBriefing.com

Chrome’s 4GB AI Model Raises Privacy Concerns in 2026 Update

Ars Technica •
×

Chrome 148 pushes a 4 GB on‑device AI model into the browser’s core, sparking confusion. The change removes a clause that previously assured users the model would stay local, raising privacy eyebrows. Many users already sought toggles to disable the feature, hoping to keep browsing private for their data privacy concerns now.

Chrome’s new toggle label claims to clarify how the browser’s APIs function, yet the omission of the “no data sent to Google” statement feels deliberate. Google maintains the model processes inputs solely on device, but the wording change, made earlier in 2026, signals a shift toward more ambiguous privacy messaging for users occasions that matter.

Users who rely on the local AI for summarizing or editing face a new reality: non‑Google sites using the API will still expose data to the site, while Google sites funnel that data to its servers. The change means that privacy guarantees depend on the website, not just the browser for every session today.

Google’s opt‑out approach forces users to scrutinize privacy policies before engaging with AI tools. As the browser remains a gateway to the web, consumers must stay vigilant: confirm whether a site’s AI routine keeps data local or forwards it to servers. Without explicit consent, anonymity is no longer guaranteed for all online activities in the.