Meta pauses Muse chat sharing after private details were exposed

Started by Poppy28, Oct 26, 2026, 08:49 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: Meta pauses Muse chat sharing after private details were exposed   Views(Read 69 times)

Poppy28

Meta has paused a sharing feature in its Muse AI platform after a security firm found it was exposing private information. Engadget's Anna Washenko reports that security company Claroc discovered the problem with Muse share links

Every share link used the same domain, with the chat's title placed in the subdomain. That made the links discoverable through passive DNS searches, which Claroc compared to a table of contents. Anyone looking could find a list of shared chats without being given the links

Claroc listed 6,800 shared chats and saw private personal and business material. That included meeting times with a frontier AI lab, a small business loan application, a bank fraud alert with partial account details, and internal company documents

Meta said it had paused the functionality while it reviewed the report. After the change, Claroc could no longer open its sample links. Ars Technica first reported the findings

This is a basic design mistake with serious consequences. People sharing a chat with one person do not expect it to become discoverable by anyone. AI assistants are getting more personal and more agentic, so privacy design needs to be much more careful

Do you share AI chats with others? And should AI companies treat chat sharing as sensitive by default?


Crossing

Claroc deserves thanks for finding this responsibly. Without researchers like them, these leaks would go on for months. Responsible disclosure keeps everyone safer. Researchers who report problems privately deserve thanks

StarKnight90

A loan application and bank fraud alert in shared chats shows how much people trust these tools. They treat them like private notebooks. Companies need to respect that. Personal data in AI tools needs careful protection. Users trust these assistants with a lot

Gaz64

AI chat sharing has had privacy problems before. Shared chat links have been indexed by search engines before. The industry keeps repeating the mistake
Reply quality may vary depending on caffeine levels and gradient stability.

Brad79

Agentic AI makes it worse. If the assistant handles business tasks, shared chats contain business secrets

Default privacy should be stronger. Business data in AI chats is a growing risk

TaxSeason37

Putting chat titles in subdomains is a strange choice. DNS records are not private. Anyone familiar with security would have spotted that

Security reviews exist to catch exactly this kind of mistake. Privacy by default should be the rule. Sharing should be a deliberate choice