Google’s New Search Setting Fuels Privacy Fears

Google is facing renewed privacy scrutiny after rolling out a new account setting that can store images, audio, video and other files people upload while using Search and related services — and, if enabled, allow some of that material to be used to improve the company’s artificial intelligence systems.

The change has drawn attention because it reaches beyond ordinary text queries into more sensitive territory: the photos people submit for reverse-image searches, recordings used in voice features and files uploaded across a growing web of Google products. For many users, the issue is not simply that Google is collecting more data, but that the training of A.I. models may be tied to everyday search behavior unless settings are changed.

A broader history setting

Google began rolling out the new control, called **Search Services History**, in June. The setting applies to signed-in users across Search, Maps, Shopping, Flights, Hotels, Translate, News and related A.I.-powered experiences. A sub-setting, **Save media**, governs whether uploaded media associated with those interactions is stored.

According to Google’s description of the feature, that history can include uploaded images, files, audio, video, voice transcripts, A.I. responses and browsing activity through A.I.-enhanced search tools. If the setting is turned on, some of the data may be used to develop and improve Google’s A.I. models.

Recent attention has centered on the practical effect of that policy: users who employ reverse-image search, Google Lens or live search features may have the media they upload saved unless they turn the setting off.

Google has described the update as a reworking of privacy controls rather than a wholly new form of surveillance, and said the rollout will take place gradually over the coming months. During that transition, some users may still see older **Web & App Activity** controls instead.

Why users are concerned

Privacy concerns have intensified because of the kind of information involved. Search uploads can include highly personal material — photos of faces, homes, documents, purchases or medical issues; audio clips; and other files submitted in moments when users may think they are making a one-time query rather than contributing to a long-term data pool.

Google says that when saved media is used for A.I. training, it is first disconnected from the user’s account. The company also says such material may be retained for up to four years. But privacy advocates have long argued that anonymization or disconnection from an account does not erase the sensitivity of the underlying content, particularly when the data includes imagery, speech and documents.

The company has also said that previously saved media may continue to be used unless a user deletes it — and that deleting it may not necessarily remove material that has already been selected for training. In practical terms, that means opting out is most consequential for future uploads, not necessarily for data already pulled into model-improvement systems.

That detail has sharpened criticism of default data-collection practices in consumer A.I. products, where the controls may exist but are often buried in account settings and discovered only after the fact.

Questions about consent and visibility

A central question is how clearly the new controls are being presented to users. Because the rollout is phased, people may encounter different settings depending on their account type, device or region. It remains unclear how many users will understand that a search interaction involving an uploaded image or voice recording could feed into A.I. development.

Google has said that turning Search Services History off stops future Search-services activity from being used to train generative A.I. models, unless a user explicitly provides feedback. The company has also said that educational accounts’ Search Services History data is not used for generative-A.I. training.

Even so, consumer advocates and regulators have increasingly focused on whether disclosures of this kind amount to meaningful consent, especially when a setting stretches across multiple services rather than a single product. Search, maps, shopping and travel tools occupy such a central place in digital life that defaults can matter more than formal choices hidden several layers deep.

Part of a larger reckoning over A.I. data

The dispute arrives as Google and other technology companies face mounting pressure over how they gather data for A.I. systems. Search companies have spent the last two years weaving generative A.I. deeper into products that were once easier for users to understand as separate from model training.

That pressure is not limited to consumers. In Britain, regulators this month announced new rules requiring clearer explanations from Google about how content is used in A.I.-driven search features and giving publishers more explicit opt-out mechanisms. The broader signal from policymakers has been that companies can no longer assume that data collected for one purpose will be accepted quietly when repurposed for another.

For Google, the stakes are unusually high because Search remains one of the company’s most widely used and most trusted products. What might once have been treated as a technical settings update is now emerging as a test of whether users believe they can still use core internet services without inadvertently helping train commercial A.I. systems.

The immediate practical effect is straightforward: users who do not want future uploads and related search activity stored for possible A.I. improvement may need to review their account settings. The larger issue, however, is whether the burden should fall on individuals to discover and disable such practices at all.

Sources

Further reading and reporting used to add context:

AI News