Google Enables 4-Year Search Media Retention for AI Training by Default

3 articles · Updated · WIRED · Jun 24

Google is rolling out a Search Services History setting that can automatically save users’ images, audio, video and files from Search interactions for AI training.
The option appeared enabled by default in testing unless users had already turned off Web & App Activity and Search Personalization, and the media-saving box was already checked.
Google says the saved material can help users revisit visual and voice searches, but a pop-up warns that if media is used to train AI models, it is detached from the account and kept for up to 4 years.
The setting covers uploads and recordings from tools including Google Lens, Search Live, Translate speaking practice and voice search, expanding Google’s data collection beyond typed queries.
Privacy advocates said the opt-out design shifts the burden onto users, arguing an opt-in model should be the minimum for a company with Google’s scale and long-standing data reach.

Sources

Left100%

WIRED2h ago

Google's New Search Settings Automatically Save User Media for AI Training

IEEE Spectrum2h ago

How to Turn Off Google Search's AI Overview - IEEE Spectrum

YouTube2h ago

Stop Google Search from Saving Uploaded Media (Stops AI Training too)

6 Sources

While rivals ask permission, why does Google default to using your data for its AI ambitions?

If deleted photos can train Google’s AI for years, do you truly control your digital footprint?

Google’s 2026 Data Retention Shift: Default Saving of Search Images & Audio for AI Training and the Privacy Backlash

Overview

As of June 2026, Google and other major tech companies treat data retention for AI training differently for enterprise and consumer users. While enterprise AI products do not keep user data for model training, consumer-facing services often retain user interactions to improve AI models. Google gives users some control, such as letting them manage chat history in the Gemini app, but details about new features like 'Search Services History'—which could default to saving user media for AI training—are not fully confirmed. This highlights the ongoing tension between advancing AI capabilities and ensuring user privacy and control.

...