Google has added a privacy setting that automatically saves user-uploaded media from Search services for artificial intelligence training. Rolling out globally over the coming months, the change means images, audio, and files shared through Google Search, Google Lens, and related tools may be stored and used to improve the company’s AI models unless users opt out.
The feature, called Search Services History, appears in Google account settings and is enabled by default for most users. It is switched off for people who previously disabled Web & App Activity and Search Personalization, but for most users, it is active without explicit consent. The option to save uploaded media for AI training is also pre-selected, so users must manually uncheck it to keep their data out of AI development.
Google describes Search Services History as a way to save past searches and activity so features such as AI Mode can personalize responses. Its scope can extend beyond typed queries: users may connect services including Gmail, Calendar, and Google Photos, allowing Search to draw on information from those accounts. Google says it does not train its generative AI models directly on a user’s Gmail inbox, Calendar, other Workspace files, or Google Photos imagery and audio. But, it may use related summaries, excerpts, generated media, and inferences to improve its services. The distinction is between using original account content directly and using material derived from it.
Google’s documentation also notes that a summary of a short or especially relevant email or file may effectively contain the whole item: “the summary may just be the email or file.” That qualification matters because a summary or inference can preserve large information even when the underlying content is not used directly. Google’s connected-app model so broadens the information that may inform Search personalization beyond what users might ordinarily consider web-search history.
Users can manage the settings through the My Activity page in their Google account, where they can review stored data, disable Search Services History, or delete past activity. To prevent image and audio uploads from being used in AI training, users must uncheck the box next to Save media. Disconnecting a connected app is a separate step: Google says it does not automatically delete information already saved in Search Services History. Previously retained history may also remain after the original email or photo is deleted. The settings description says that once data is used for this purpose, it is separated from the user’s account and may be retained for up to four years, even if the original activity is deleted.
Google says the settings are intended to help users revisit previous searches, including visual and voice queries, and can be adjusted at any time. The settings page lists typed queries, images, files, and audio or video recordings from interactions with Search services, including Google Lens uploads, voice searches, and recordings from Search Live or Translate speaking practice. The controls distinguish among turning off future saving, disconnecting an underlying service, and deleting historical Search Services History.
The expansion reflects how AI personalization can draw on different forms of user activity and connected content. Google’s consumer-account policies differ from its Workspace commitments: for personal accounts, activity may be used to improve services, including generative-AI training, when the relevant activity-saving setting is enabled. Google says Workspace customer data, including prompts, is not used to train models without the customer’s prior permission or instruction. These separate policies put consumer settings and enterprise data under different terms.
Critics argue that requiring users to opt out, rather than asking them to opt in, puts the burden of protecting privacy on individuals. They say companies should seek explicit consent before enabling features that use personal data for AI development. Google presents the change as giving users more control over saved history and points to conveniences such as revisiting past visual searches or continuing conversations. Still, the design makes it impressive for users to review both connected-app permissions and Search Services History itself.
In Europe, the issue is part of a wider regulatory debate over personal data and AI development. In September 2024, Ireland’s Data Protection Commission opened a cross-border inquiry into whether Google should have conducted a data-protection impact assessment under Article 35 of the General Data Protection Regulation before processing European personal data to develop its PaLM 2 model. Article 35 requires such an assessment when processing is likely to create a high risk to people. The inquiry concerns Google’s broader practices for AI-training data, not a final decision specifically about Search Services History. The GDPR’s transparency and data-protection requirements apply to consumer services across Spain and other EU countries. The inquiry also illustrates the separate regulatory scrutiny surrounding the use of personal data in AI development.
