Google Chrome may be eating up several gigabytes of storage on the user’s device without their realizing it, all because of artificial intelligence. The announcement has led to a debate on transparency, privacy, and user consent, as it is reported that Google has been secretly installing Gemini Nano, its on-device AI model, on some devices.
Gemini Nano is a light version of Google’s Gemini AI that will be used on smartphones, laptops, and other devices without the need for cloud-based processing. The model is used in a variety of AI features such as detecting scam calls, message composition assistance, content summarization, and supported device image analysis.
Reports indicate that Chrome might automatically install the AI model if the device has certain hardware capabilities. Even though the installation takes place automatically, and no user action is required, many users may not know that the model has been installed on their system.
The size of the model is one of the main points of concern with the issue. Gemini Nano will take up around 4GB of storage space, which could be a big concern for those with limited storage. Google claims that the model is only installed on a device that has enough resources, but critics say users should have known prior to downloading such a large file.
Google has explained that the model will automatically unload itself if the device doesn’t have sufficient processing power, memory, storage space or network resources. Also, the users can now disable and remove the model directly from the Chrome settings, the company adds. Once the feature is disabled, subsequent model updates and downloads will stop, Google says.
The revelation has sparked discussions about user consent and transparency in the application of AI technologies more generally. Privacy rights groups say that software companies need to be more transparent about when and how big AI models are being installed on users’ machines, particularly when the models are using up storage, computing power and energy.
However, there have been indications that such practice could create concerns for regulation in areas that have stringent privacy frameworks. There have been concerns regarding the transparency and consent aspects of automatic deployment of AI models in line with the data protection laws, especially in the European Union.
On-device AI may be beneficial for Google in a few ways. Local execution can boost response times, ensure better data privacy and reduce the costs of operating extensive AI systems. Some AI capabilities can also be performed locally, which can help to minimize reliance on internet connections.
You can check for the presence of Gemini Nano by looking into Chrome data directories on the user’s device manually. If the model exists, it will be an “OptGuideOnDeviceModel” folder with AI model files. Google has also made settings available in Chrome to turn off on-device AI tools and block future downloads.
The case illustrates a new dilemma that AI is bringing to the tech sector, as it increasingly infiltrates commonplace software. AI-driven improvements can provide valuable functionality and enhance user experiences, but manufacturers should ensure that the user is aware of what is being installed in their devices and how this technology works.
As AI progresses from cloud systems to individual devices, it is likely that the role of user awareness, informed consent, and good communication will become more significant in fostering trust between technology providers and users.
