Google has expanded its artificial intelligence training data to include media files uploaded by users through various search-related services. This update covers images, audio, video, and other files submitted using tools like Google Lens and Google Translate, broadening the pool of information used to improve its AI models.
The change, implemented quietly, enables Google to harvest data from visual identification requests, voice-activated searches, and uploads processed via Google Translate. Users’ photos, audio clips, and videos automatically become part of the training datasets unless they actively opt out.
To exclude their media from AI training, users need to access the Search Services History settings and uncheck the “Save Media” option. In addition, reviewing Search Services Personalization controls can prevent further retention of uploaded files for AI purposes. For those who wish to avoid AI-generated search results entirely, appending “-AI” to queries removes such content from their search interface.
This policy currently applies only to search-related products and excludes personal storage services like Google Photos. The decision highlights the escalating demand for fresh, high-quality data in generative AI development, as companies race to enhance their language models through unique user interactions.
Google’s approach reflects a wider strategy in the technology sector to leverage user-generated content to sustain competitive advantages in AI. This follows earlier instances where Google sought permission from news publishers to utilize their content for AI training or face financial penalties related to Google News.