Loading...
Loading...
Google’s Chrome has quietly removed a statement asserting its on-device AI never sends user data to Google servers, fueling concerns about data handling transparency. The change suggests Chrome may transmit certain inputs or telemetry off-device for features like personalization, error analysis, or model updates. Privacy advocates warn users to expect some data flows despite earlier assurances, and experts urge clearer documentation and opt-out controls. The shift reflects broader tensions as browsers add AI features: balancing functionality, model improvement, and privacy requires explicit disclosures about what is processed locally versus what is uploaded to company servers.
Google Chrome is downloading a 4GB on-device Gemini AI model for local processing, but this isn’t a new rollout — Chrome has included that local model for years. The recent flurry of attention came after users spotted a 4GB Gemini Nano model and assumed Google had suddenly deployed AI across all Chrome installs. In reality, Google announced in 2024 it would add local AI capabilities to power features like Help Me Write, tab organization and scam detection; the company’s poor communication about those background downloads and storage needs caused the confusion. This matters for user trust, device storage management and transparency around browser-integrated AI.
Chrome reportedly auto-downloads a ~4GB on-device AI model, likely Google’s Gemini Nano, to enable local features such as text summarization. The user asks two questions: the exact model name/version Chrome installs, and whether a GGUF-format file is available to run locally with llama.cpp. This matters because an on-device model raises privacy, storage, and portability considerations, and GGUF would let users run the model offline with open-source tools. Key players are Google (Chrome, Gemini Nano) and the open-source runtime ecosystem (llama.cpp/GGUF). There’s no definitive public confirmation of the exact Chrome model build or an official GGUF release; availability depends on Google’s packaging and licensing decisions.
Chrome on desktop includes a roughly 4 GB on-device Gemini Nano AI model that auto-downloaded for many users after Google integrated it to enable local AI features like scam detection and developer APIs. Reports that users were unaware of the file prompted attention; Google added an “On-device AI” toggle (rolling out since February) in Chrome Settings → System to disable and remove the model, and says disabling stops future downloads and updates. Google and Chrome GM Parisa Tabriz argue on-device processing preserves privacy and powers security features, while critics warn users may miss such significant changes and that disabling Gemini Nano will disable local security and developer functionalities.
Chrome removes claim of On-device Al not sending data to Google Servers
Chrome removes claim of On-device Al not sending data to Google Servers