Google Trains AI on Search Photos and Recordings by Default Unless Users Opt Out

Google has started using media that users upload through its search services, including Lens, Translate, and voice search, to train its AI models. The change took effect quietly, and every user is enrolled in it by default.

What Google is collecting

The data in question is media that users themselves upload while using search tools: a photo uploaded to Google Lens to identify an object, a voice recording accompanying a spoken search, a document submitted to Google Translate for translation. All of it can be used to develop the company's AI models and safety systems.

The scope of the change extends beyond search itself. It also covers Maps, Shopping, Flights, Hotels, and Google Messages, services where users regularly upload content that has nothing to do with typing a conventional text query.

Default opt-in and how to turn it off

The key problem commentators point to is how the change was rolled out. Users were enrolled automatically, without being asked for consent, and notice of the change arrived mainly as an email that is easy to miss.

Similar to your Search history, saved media is also used to develop and improve Google services and technologies, including AI models and safety mechanisms - Google notice to users

To opt out, users need to go to the Search services history page and uncheck the Save media option separately from their overall search history. On the search personalization settings page, they can also shorten the data retention period to 3 months instead of the longer defaults. Some guides also point to appending -AI to queries to limit AI-generated results in search itself, though that is a separate matter from model training.

The AI industry's data hunger

Google's decision fits a broader trend. Companies developing large language and multimodal models are increasingly feeling a shortage of fresh, high-quality text, voice, and visual data after exhausting the internet's easily available resources. Earlier in 2026, Google had already pressured news publishers to make their content available for AI training, under threat of losing annual payments from Google News.

Meta takes a similar approach to photos and content posted by users of its platforms, showing that defaulting users into AI training data collection is becoming standard practice across the industry rather than an isolated move by a single company.

What this means for users in Poland

For people and businesses in Poland who regularly use Google Lens to identify products, Google Translate to translate documents, or voice search on their phones, this means uploaded files may end up in the AI training data pool unless someone manually changes the settings. That also applies to businesses that upload documents containing sensitive or trade-secret information through these tools.

From the perspective of GDPR and EU data protection rules, defaulting users into processing for AI training purposes, without explicit opt-in consent, could raise questions for regulators, especially amid ongoing work to clarify the AI Act's rules on training data.

For now, the only way to protect against this is to review Google account settings manually and turn off media saving, which requires users to first be aware that the option exists at all.

Sources: Google now uses your uploaded search media to train AI (engadget.com), If you use Google, you're training its AI. Here's how to opt out (techcrunch.com)