Hidden 4GB AI file in Chrome triggers privacy backlash against Google
A hidden 4GB artificial intelligence model allegedly being installed by Google through its Google Chrome browser has triggered criticism from computer scientist and lawyer Alexander Hanff, who claims the practice may violate privacy and environmental standards, reports a Qazinform News Agency correspondent.
In a blog post, Hanff said Chrome has been “reaching into users’ machines and writing a 4GB on-device AI model file to disk without asking.” The file, known as “weights.bin,” is reportedly linked to Gemini Nano, Google’s on-device large language model.
According to Hanff, the file is installed automatically when Chrome’s AI features are enabled, which he says are turned on by default in recent browser versions.
“Chrome did not ask. Chrome does not surface it. If the user deletes it, Chrome re-downloads it,” Hanff wrote.
He argued that users are given no visible consent option for downloading the model, adding that removing the file requires several undocumented steps.
The researcher also criticized what he described as a broader industry trend of deploying AI tools directly onto personal devices without clear approval from users.
“An engineering team at a large AI vendor decided that the user’s machine is a deployment surface to be optimized for the vendor’s product roadmap, not a personal device whose owner is the legal authority on what runs there,” he said.
Hanff warned that the environmental impact could also be significant due to Chrome’s massive global user base. He estimated that distributing the model across devices could generate between 6,000 and 60,000 tonnes of CO2-equivalent emissions.
“That is the environmental cost of one company unilaterally deciding that two billion people’s default browser will mass-distribute a 4GB binary they did not request,” he added.
Hanff suggested Google should instead introduce an explicit opt-in request before downloading the AI model onto users’ devices.
Earlier, Qazinform News Agency reported that the United States Department of War had announced a series of agreements with leading artificial intelligence companies to integrate advanced AI capabilities into its most secure classified networks.
