The privacy debate around artificial intelligence has intensified after recent changes by Google and Meta highlighted how user data can be used to improve AI models. The issue is no longer limited to search queries. AI companies are increasingly interested in the data users create while interacting with their services — chats, photos, voice recordings, documents and search activity.Why has this become a bigger issue now?■ Google began rolling out its new Search Services History settings on May 26, separating search activity controls from the older Web & App Activity settings. Under the new system, images, audio recordings, videos, and files used across services such as Search, Maps, Translate, News, and Lens may be stored and potentially used to improve AI systems unless users turn the option off. Google said the rollout would continue over the following months.Meta’s AI photo tools drew attention in the first week of July 2026, when the company expanded features, including Muse Image, which can create AI-generated images using eligible public Instagram photos unless users opt out.What are Google and Meta saying?■ Google has defended the changes as giving users “more control” over their search history and personalisation settings. The company says users can disable Search Services History entirely or choose whether images, files and audio recordings are saved. “We take your privacy seriously and take steps to protect it when using your Search Services History data to train our AI models,” said Google in a support document.Meta says its AI features are intended to help users create and edit images and that people can manage eligible AI settings through the privacy controls available on its platforms. “At any time, you can turn off or allow reuse for reels, feed videos and photos across your entire account,” said Meta.Which platforms may use your data?■ The broader issue extends beyond Google and Meta. OpenAI’s ChatGPT allows many consumer-account conversations to be used for model improvement unless users disable the relevant training settings. Microsoft Copilot collects prompts, documents and usage data to improve services, although commercial customers receive stronger protections. Anthropic’s Claude and xAI’s Grok can also use some consumer interactions and uploads for model improvement, subject to their privacy settings and opt-out options.
What data these AI platforms use
Does paying for a service stop data from being used?■ Not necessarily. Consumer subscriptions mainly provide additional features and higher usage limits. The strongest restrictions are usually found in enterprise, education and commercial plans, where providers make contractual commitments not to train models on customer data.What should users check?■ Privacy specialists say users who are concerned about AI training should review their settings rather than assume defaults are restrictive. On Google services, that means checking Search Services History, Web & App Activity and auto-delete settings. ChatGPT users can disable “Chat History & Training” if they do not want conversations used for model improvement.Meta users should review Instagram and Meta AI privacy controls and opt out of AI features that use public photos where available. Claude users can check data controls in settings and disable model-improvement participation where offered, while Grok users should review privacy settings on X and avoid sharing sensitive personal information.
