Meta Pauses Muse Sharing Feature After Security Firm Discovers Publicly Available AI Chats
Meta has paused a feature in its AI assistant, Muse, that allows users to share conversations via links after researchers found these URLs publicly accessible on the web. A security firm named Claroc discovered over 6,800 such shared Muse links through public DNS records and reported them as easily discoverable due to how Meta structured the link names. The affected feature lets users generate a unique URL for their AI conversations that can be accessed by anyone with the link. However, because these URLs are built from predictable titles and IDs, they became searchable via DNS data. Claroc found sensitive information in some of these chats, including business revenue details, loan applications, staff calendars, and job interview notes. This incident highlights ongoing concerns about privacy risks associated with AI chat features across platforms like OpenAI’s ChatGPT and Anthropic’s Claude.
Tickers: $META
Topics: cybersecurity data privacy information leaks Meta Muse AI