Google Quietly Expands AI Training Data Collection, Defaults to Opt-In

Google has quietly rolled out significant changes to its privacy settings, allowing its AI models to be trained on a broader spectrum of user data, including uploaded images, files, audio, and video from various Search services. This update, currently deploying globally, introduces a new "Search Services History" setting that, by default, enables the collection of media shared through tools like Google Lens, Google Translate's speaking practice, and voice searches.

This move marks a substantial shift in Google's data utilization strategy for AI development. While Google maintains that data used for AI training is "disconnected" from user accounts and filtered for sensitive information before human review, the default opt-in places the onus on users to actively manage their privacy. The collected data, crucial for enhancing the accuracy and performance of Google's AI, particularly its Gemini models, may be retained for up to four years, even if the original activity is deleted by the user.

For users concerned about their data contributing to Google's AI, opting out is possible, though it requires navigating updated settings. To prevent Google from using your media for AI training, users should visit their Google Account's "Data & Privacy" section, then "Activity Controls," and locate "Search Services History." Within this new section, the "Save Media" option can be unchecked. If this setting is not yet visible, users can manage media-related data collection by turning off "Include voice and audio activity" and "Include Visual Search History" under their existing "Web & App Activity" controls. This development underscores the ongoing tension between technological advancement and individual privacy, urging users to remain vigilant about their digital footprints.