Google expands AI training to all uploaded media unless users opt out

Google has automatically enrolled all users in its expanded AI‑training program, allowing images, audio, and video uploaded to Search services to be used for large‑language‑model training unless the user opts out [1].

Expanded media scope

The updated settings, announced in an email last month, add voice searches, screenshots, and any media you submit to Google’s search tools to the “Search Services History” pool. When “Save media” is enabled, Google may use that content to improve its AI models and related services [1]. The policy applies to all signed‑in users, meaning the default behavior is to retain and repurpose personal media unless the user manually turns the feature off.

How to control the data flow

Google provides two settings pages—Search Services History and Search Service Personalization—where users can uncheck “Save media” or disable the entire history feature. The controls are also accessible via the My Activity portal (myactivity.google.com) and allow users to set retention periods of 3, 18, or 36 months [1]. Disabling the feature stops the data from being fed into the training pipeline, but keeping it on can enhance personalized search and ad experiences.

What to watch

• Google’s rollout timeline – Monitor any further updates to the policy or additional media categories that may be added.
• Regulatory response – Watch for potential privacy regulator inquiries or new legislation targeting AI‑training data practices.
• User opt‑out rates – Track how many users disable the “Save media” option, which could signal broader privacy concerns.

The move underscores Google’s push to enrich its AI models with real‑world user content, while raising fresh privacy questions for anyone who relies on Google’s search ecosystem.