Tech firms track engagement to train AI models

Google's new Search Services History settings (rolling out from May 26, 2026) may store images, audio, videos, and files from Search, Maps, Translate, News, and Lens for AI training unless users opt out.

Meta expanded AI photo tools in July 2026, including Muse Image, which can generate images using eligible public Instagram photos unless users opt out.

OpenAI's ChatGPT, Microsoft Copilot, Anthropic's Claude, and xAI's Grok may use consumer interactions and uploads for model improvement subject to privacy settings.

Paying for consumer subscriptions does not guarantee data is excluded; strongest protections are in enterprise, education, and commercial plans with contractual commitments.

Users should review privacy settings on each platform to disable data use for AI training where opt-out is available.