Google now uses your images and voice to train AI models

Google has silently expanded its AI training data to include all media you use in search-related interactions — images, videos, and voice recordings. As of last month, Google automatically opted all users into this policy unless they manually disable it. The company updated its Search Services History documentation to state that saved media (images, files, audio, video) may be used to 'develop and improve Google's AI models and technologies.' This extends to media uploaded via features like Circle to Search, voice searches on Search Live, and even language practice with Google Translate. The privacy risk is significant: your voice could be used to train LLMs, and sensitive conversations (e.g., discussing leaving a job) could inadvertently reappear in responses to others.

To protect your privacy, you must take action. Open the Google app, tap your profile image, then go to Search history. Uncheck 'Save media' and turn off the Search Service History toggle. Then, go to Search personalization and disable that service as well. These steps prevent future media from being saved for AI training. While Google frames this as improving search and AI experiences, the default opt-in leaves users exposed. As AI becomes more embedded, users need to stay vigilant about privacy controls — especially when companies quietly shift policies to use personal data for model training.