Ticker

6/recent/ticker-posts

Meta AI app leaks users' highly private conversations

Meta AI app leaks users' highly private conversations

Since the launch of the Meta AI app last April, its public feed has been full of very private discussions between users and the chatbot. Unlike ChatGPT or other AI apps, Meta has chosen to create a public space where users can share an image or a query. The problem is that users tend to allow sharing a little too easily.

A Public Feed That's a Little Too Intimate

The result: all sorts of conversations can be read in Meta AI's public feed, ranging from the trivial (prayers, poems) to the downright embarrassing (requests for medical advice, desperate pleas for help, etc.). Even more worrying: audio conversations mixing user and bot voices were accidentally shared.

Meta AI app leaks users' highly private conversations

These messages, some of which are very personal, are exposed to all users of the app. Meta explained to Business Insider that several steps were required to make a conversation public. However, the app's interface appears ambiguous enough to mislead users.

Faced with the embarrassment of this public exposure, Meta has implemented a first protective measure: an explicit warning now appears before each attempt to publish. This new pop-up appears when the user taps "Share": "[Meta AI] queries you post are public and visible to everyone. Your prompts may be suggested by Meta in other Meta apps. Avoid sharing personal or sensitive information."

After this message, you must validate the publication a second time. At the same time, Meta has also modified the public feed to display only AI images, which seems less problematic than exposing marital or child custody issues.

The underlying problem, however, remains: many users do not fully understand the implications of what they are sharing, or do not imagine that their conversation with a chatbot could be visible to everyone.

Source: Business Insider

Post a Comment

0 Comments