Meta pauses Muse AI share-link feature after researchers expose private chats
Security researchers at Claroc found they could access personal conversations shared through Meta's Muse AI by querying passive DNS records, since the service embeds each conversation's title directly into a shareable subdomain. Meta confirmed it has paused the share-link functionality while reviewing the report and considering changes to how the links are discovered, after Claroc reported the issue through Meta's bug bounty program on October 8.
GoKawiil's interpretation of the reporting above, not reported fact.
The flaw suggests that design choices around how share links are structured can inadvertently make private AI conversations publicly discoverable with minimal technical effort, since passive DNS tools are widely accessible to researchers and potentially bad actors alike. This episode could prompt broader scrutiny of how AI chat products handle shareable links, especially as such tools increasingly process sensitive personal or financial information.
- Claroc researchers found Muse share links could be found via a single passive DNS query, exposing private conversation titles and content.
- The vulnerability stemmed from embedding conversation names directly in the shareable subdomain rather than hiding them in a URL path.
- Meta has paused the share-link feature while it reviews the report and considers redesigning how links are generated and discovered.
Source: wired.com — Dan Goodin, 2026-10-21
Published there as: “A Design Flaw in Share Links Exposes Users’ Personal Conversations With Meta’s Muse”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.