Your AI Chatbot May Be Leaking Your Chats to Meta, TikTok and Google
A study by IMDEA Networks Institute found that major AI chatbots—ChatGPT, Claude, Grok, and Perplexity—share user data with third-party advertising and analytics platforms like Meta, Google, and TikTok through embedded trackers. Over 13 trackers were identified, none transparently disclosed to users. These services send chat URLs and user identifiers, sometimes even message content, to tracking companies. Some platforms, such as Grok, make guest chats public by default, while Claude and ChatGPT offer stronger controls but still transmit metadata. Claude routes data via its servers, bypassing ad blockers. Perplexity recently removed its Meta tracker. While there is no evidence that companies like Meta or Google are directly reading conversations, the technical ability to do so exists, posing privacy risks. Users can limit exposure by adjusting privacy settings and rejecting cookies, but full protection is difficult. Researchers plan to assess additional AI platforms in future studies and have reported their findings to regulatory authorities; none of the involved companies have responded so far.
