Meta Muse Shared Links Leaked Users' Private Chats
Security firm Claroc found more than 6,800 open links to conversations with Meta's AI assistant Muse. Some contained users' personal and business secrets.
Security firm Claroc found more than 6,800 publicly accessible links to conversations with Muse, Meta's AI assistant. Each link is built from the name of the conversation plus a random code on the domain cf.metaaiusercontent.com. Claroc found them through a simple passive DNS search, which looks at records of past requests to the domain. The researchers did not guess any codes. One of the exposed conversations contained a user's to-do list, including a warning about bank fraud and part of a bank account number. Meta paused the sharing feature after being notified.
The problem is not the AI model itself. It is a common web design mistake. A predictable web address, instead of a private, hard-to-guess link, turned personal conversations into pages that search engines and DNS archives could find. Any business that gives customers a similar share button in an AI chat should check whether its own links are just as easy to find.
Source: arstechnica.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides

Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles