Google Trains AI on User-Uploaded Search Media by Default, Covering Images, Audio and Video
Updated
Updated · Engadget · Jul 6
Google Trains AI on User-Uploaded Search Media by Default, Covering Images, Audio and Video
3 articles · Updated · Engadget · Jul 6
Summary
Google has quietly begun using media uploaded through Search-related tools to train its AI models, with users automatically enrolled unless they change privacy settings.
That policy covers images, files, audio and video from products such as Google Lens, voice search and Google Translate, while personal Google Photos are not included.
Users can opt out by turning off “Save Media” on the Search Services History page and disabling saving on the Search Services Personalization page.
The change reflects a broader push for more training data as generative AI companies expand collection of user-submitted content across consumer services.
How does Google's default media collection comply with strict global privacy laws?
You hit delete, but does Google's AI keep your data for four more years?
Google's June 2026 AI Data Collection Overhaul: Multimodal Inputs, User Privacy Controls, and Regulatory Challenges
Overview
In June 2026, Google expanded its data collection practices for AI training by introducing the 'Search Services History' setting, which now manages not only traditional search data but also visual and audio inputs. This update supports Google's push to develop advanced multimodal AI features like AI Overviews and AI Mode, which rely on processing a wider range of user interactions. By broadening the types of data collected, Google aims to enhance its AI's reasoning and capabilities in Search. The new setting gives users more control over their data, reflecting Google's strategy to balance innovation with user privacy.