How to Opt Out of Google Search’s New AI Data Training Feature

Google is rolling out a new setting that automatically saves users' Search history media — including images, audio, and video — to train its AI models, requiring users to opt out if they object to the data collection. The feature, called Search Services History, was already enabled and the box to save uploaded media was pre-checked when a WIRED reporter visited the settings page [1]. A Google spokesperson, Davis Thompson, said the settings “help users get more relevant results and revisit their searches—including visual and voice searches—and they can be turned on or off at any time” [1]. He did not answer a question about the feature being on by default [1]. Google will store saved media used for AI training for up to 4 years, even if the original activity is deleted [1]. The company stated that media used to train its AI models “is disconnected from your Google Account” [1]. Users can opt out by visiting the My Activity page and selecting the Search Services History tab, then unchecking the box next to “Save media” [1]. The data collection spans interactions across Google’s ecosystem. The company’s Gemini family of AI models is trained natively on multiple data types, allowing the models to process and generate text, code, images, audio, and video simultaneously [2]. Google distributes Gemini in versions ranging from on-device “Nano” to high-compute “Ultra” models, and the technology integrates into services including the Gemini mobile app and the Vertex AI platform for developers [2]. Thorin Klosowski, a senior security and privacy activist at the Electronic Frontier Foundation, said Google’s position is distinct because it offers so many services that people have used for a long time and “have grown pretty comfortable and complacent with the amount of data collected” [1]. Klosowski argued that opt-in consent “is really asking the bare minimum of these companies” [1]. Ben Winters, director of AI and privacy at the Consumer Federation of America, said the change “creates this extra layer of math that a consumer has to do about whether they feel comfortable using the tool they’ve been using for a long time” [1]. He described “an increasing feeling of powerlessness and hopelessness about even trying to protect your data, because every little thing is going to be squeezed out of you” [1]. Google’s approach mirrors a broader industry pattern. OpenAI’s Sora text-to-video model, for instance, used copyrighted material by default unless copyright holders actively opted out [4]. Google has faced sustained criticism over privacy practices, with regulators in Austria, France, and Italy finding that certain uses of Google Analytics violated the General Data Protection Regulation due to data transfers to the United States [5]. The company has also drawn scrutiny for its compilation of user data and its monopoly position across search, advertising, and mobile operating systems [6].