AI assistants leak chat data to advertisers, study finds

A team of researchers from IMDEA Networks and Carlos III University of Madrid has conducted a study revealing that AI assistants, including ChatGPT, Gemini, and Claude, may not be as confidential as users believe. By analyzing nine AI tools in both web and mobile app versions, they found that all of them connect to at least one advertising or tracking service, raising serious questions about the privacy of user conversations.

This finding is particularly significant at a time when AI assistants have become our digital confidants, where we share everything from work issues to personal matters. The trust placed in these tools is now threatened by the possibility that user data may be used without their knowledge by advertisers.

## Why it matters Privacy in the use of AI technologies is an increasingly concerning topic. The revelation that user data may be leaked to third parties raises ethical and security questions about the design of these applications. Users need to be aware that their interactions may not be as private as they believe.

## What we know The study was presented at the Privacy Enhancing Technologies Symposium (PETS) and encompasses a thorough analysis of various AI platforms. Six out of the nine tools analyzed have been identified as data transmitters to advertising services, suggesting that the majority of AI assistants do not guarantee the confidentiality that many users expect.

## What is still unclear The exact extent of the data being leaked and how it is used by advertisers has not yet been confirmed. The companies have not detailed what types of information are shared, leaving users with more questions than answers about the security of their interactions.