Chrome installed a 4GB AI model without warning

Google first announced the integration of the Gemini Nano model into Chrome in 2024, while the rollout of local AI features began with Chrome version 126. Support was later expanded to more devices and additional capabilities, at a time when artificial intelligence is becoming an increasingly common part of everyday internet use.

However, the company has come under criticism after it was discovered that Chrome had automatically downloaded an AI model of nearly 4 GB onto some users’ computers, without clear consent or sufficient notice about what was being installed.

The model in question is Gemini Nano, which allows certain AI features to run directly on the device, without requiring every request to be processed on Google’s servers. Users, however, were not clearly asked whether they wanted to download such a large model, nor were they properly informed about when the download begins, which features use it, or how it can be disabled or removed.

Critics warn that this approach could normalize the installation of large AI components without sufficiently clear and visible user consent. This raises further concerns about transparency, privacy, and users’ control over the software installed on their devices.