Google Now Stores Your Search Media to Train Its AI Models

The News

Google has quietly reworked its privacy controls so that the images, audio clips, files and video you hand to its search products can be retained and used to train its AI models. The change surfaced through a customer email in June 2026 and was detailed by TechCrunch on 6 July 2026.

The old unified "Web & App Activity" toggle has been split into two independent controls: Search Services History and Personalised Recommendations. Sitting alongside them is a separate Save Media option, switched on by default, that governs whether Google keeps the media you generate while searching.

That media net is wide. It covers pictures run through Google Lens, voice recordings from voice search and the Search Live feature, audio from Google Translate practice sessions, and files or video uploaded across Search, Maps, Shopping, Flights, Hotels, Translate and News. In its own words, Google says saved media "is also used to develop and improve Google services and technologies, including AI models and safety measures."

Users can set saved data to auto-delete on a 3, 18 or 36 month cycle, but the retention is opt-out rather than opt-in.

Why It Matters

The significance is less about any single setting and more about the pattern. By breaking one clear toggle into several granular ones and leaving media capture on by default, Google shifts the burden of privacy onto the user who must now hunt through menus to say no. It is the same design tension that surfaced when Meta began using public posts to train its models, and when Zoom faced a backlash in 2023 over terms that appeared to permit training on customer content.

What is different in 2026 is scale and sensory range. Training data is no longer just text and clicks. It is your voice, your uploaded documents, the photograph you snapped to identify a plant or a medicine strip. As frontier labs compete for multimodal training material, everyday search behaviour becomes a renewable data mine, and default settings become the quiet lever that fills it.

Indian Angle

For India, this collides directly with the Digital Personal Data Protection Act, 2023, whose rules the government has been finalising through 2026. The Act leans heavily on the idea of informed, specific consent and gives users the right to withdraw it as easily as they gave it. A media-capture setting that is on by default, buried under a renamed menu, is exactly the kind of design that Indian regulators at MeitY have signalled they want to scrutinise once the DPDP framework is fully enforced.

There is also a competitive read for Indian model builders. Homegrown efforts such as Sarvam and Krutrim are chasing Indic multimodal data, and Google's default retention shows how much of an advantage a consumer platform with hundreds of millions of Indian users enjoys in sourcing voice and image data across dozens of Indian languages. For a Bengaluru startup, that data moat is harder to replicate than the model architecture itself.

For Indian enterprises and developers, the practical takeaway is governance. Teams using Google search tools inside workflows should assume uploaded files may be retained, review the Save Media setting on corporate accounts, and factor DPDP consent obligations into any customer-facing product that pipes user media through these services.

FAQ

When did this change take effect?

Google notified users of the updated privacy settings via a customer email in June 2026. TechCrunch published its detailed explainer on 6 July 2026. The Save Media option is already active by default on affected accounts.

What exactly is Google storing now?

Images from Google Lens, audio from voice search and Search Live, audio from Translate practice sessions, and files or video uploaded across Search, Maps, Shopping, Flights, Hotels, Translate and News. This saved media can feed AI model development and safety systems.

How do I opt out?

Open the Search Services History settings page and uncheck the "Save Media" box separately from Search Services History itself. You can also set data to auto-delete every 3, 18 or 36 months, and find further controls at myactivity.google.com/more-activity.

Does India's DPDP Act cover this?

The Digital Personal Data Protection Act, 2023 emphasises specific, informed consent and easy withdrawal. Default-on media retention could draw regulatory attention once the rules are fully enforced, though enforcement specifics are still being finalised by MeitY.

Where can I read the original coverage?

The full report is on TechCrunch, linked in the attribution note below.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.