Google defends 4GB background AI downloads amid privacy backlash

Google Chrome VP and GM Parisa Tabriz responded to criticism over Chrome’s silent 4GB AI model downloads, emphasizing that on-device AI is vital for the browser’s security and development.

Privacy researcher Alexander Hanff documented that Chrome automatically downloads the Gemini Nano model, about 4GB, onto devices without user prompts or clear opt-out options.

Deleting the file causes it to re-download upon Chrome restart. The controversy grew when critics pointed out that Chrome’s visible “AI Mode” feature in the address bar doesn’t use the local model.

Instead, it sends queries to Google’s cloud, meaning users bear the cost of a 4GB file unrelated to the main AI feature. Privacy advocates also expressed concern over possible violations of EU privacy laws, notably the requirement for user consent before local data storage.

Tabriz clarified on X that Google has been offering Gemini Nano as a lightweight, on-device model since 2024, essential to Chrome’s APIs and security features like scam detection. She explained that the model processes data locally and uninstalls itself if storage is low.

However, she did not directly address consent concerns or explain why the model reinstalls when deleted. Google stated that users can disable or remove the model via Chrome settings, and it won’t re-download once turned off.