Google Chrome users may have an unexpected 4GB file sitting on their computers as part of the company’s rollout of on-device artificial intelligence features.
The file, known as Gemini Nano, is a compact AI model that runs directly on users’ devices rather than relying entirely on cloud servers. According to privacy researchers, has been quietly downloading the model onto eligible computers without displaying a clear notification or consent prompt.
Gemini Nano powers a growing range of AI-powered capabilities, including text assistance, content summarisation, scam detection, and other features designed to work locally on a device. Google says processing tasks on-device can improve speed, reduce reliance on internet connectivity, and enhance privacy for certain functions.
However, the installation has sparked criticism among privacy advocates, who argue that many users are unaware that the software has been downloaded. The AI model is stored as a file called weights.bin, which can occupy around 4GB of storage space.
Google says the model is installed only on devices that meet specific hardware requirements and may be automatically removed if the system lacks sufficient storage, memory, processing power, or network resources. The company has also introduced settings that allow users to disable on-device AI functionality.
Mac users can open Finder, navigate to Library &; Application Support > Google > Chrome > Default, and look for a folder named OptGuideOnDeviceModel. If the folder contains a file called weights.bin, Gemini Nano is installed.
Windows users can press Windows + R and enter %LOCALAPPDATA%GoogleChromeUser DataOptGuideOnDeviceModel. Alternatively, they can browse to the same location through File Explorer and check for the weights.bin file.
To disable Gemini Nano on a Mac, users should open Chrome, go to Settings > System, and switch off On-device AI.
Windows users can follow a few additional steps:
The controversy highlights a broader shift in the AI industry. Technology companies are increasingly moving AI processing from cloud servers onto users’ own devices, reducing infrastructure costs while enabling faster and more personalised experiences.
Critics argue that such deployments should be more transparent, particularly when they involve downloading large files and using local storage without explicit user approval. Some privacy experts have also questioned whether the approach complies with transparency requirements under European data protection laws.
While Google maintains that Gemini Nano is intended to improve Chrome’s AI capabilities, the episode has reignited debate over how companies introduce AI features and the level of control users should have over software installed on their devices.
