AI companies leak data to advertisers [pdf] (jorgegarciaherrero.com)

A recent privacy analysis conducted by researchers from IMDEA Networks revealed that popular conversational AI services, such as ChatGPT, Gemini, and Claude, are increasingly leaking user-generated data to third-party advertisers and tracking services. The study systematically reviewed nine AI platforms, uncovering that all of them integrate at least one third-party tracking or advertising service. This creates significant privacy risks, as these platforms not only handle sensitive user interactions but also disclose unique conversation-derived artifacts, including prompts, titles, and even entire conversations, often without sufficient access controls or user consent. The findings raise critical concerns for the AI/ML community as this integration of traditional tracking frameworks with conversational AI can expose users to new privacy invasions. Data-sharing practices could lead to increased profiling and targeted advertising, undermining users' expectations of confidential interactions with AI. As these technologies become more pervasive, there is an urgent need for robust regulations and stronger safeguards to ensure transparency and protect user privacy in AI-mediated communications, particularly in light of existing frameworks like the GDPR.