News

Policy & safetyOct 20, 20263 min read1 source

Meta’s Muse AI agent leaked shared chats containing business and personal details

Act now

Researchers found Muse share links could be discovered by outsiders, exposing loan applications, revenue reviews and more. Meta has paused sharing.

Read it in

Drafted with AI· Oct 20, 2026

Policy & safety · Oct 20, 2026

The distilled version

  • Security researchers found that shared Muse conversation links could be discovered by people who were never given the link, exposing sensitive business and personal data.
  • Exposed content reportedly included a loan application, a company revenue review, a bank fraud alert, and a staff vacation calendar.
  • Meta has paused the sharing feature while it reviews the issue. Treat any content you’ve shared via a Muse link as potentially exposed.

What happened

Ars Technica reported on October 20 that security firm Claroc found Meta’s new AI agent Muse let outsiders discover shared conversation links without ever being sent them. The researchers said they enumerated 6,800 share links and found sensitive content in a sample of a few hundred, including a candidate’s notes for an executive-search call, a marketing company’s internal revenue review, a small-business owner’s loan application, a bank fraud alert with partial account details, a real-estate investor’s property performance report, and a company’s full-year staff vacation calendar.

The problem stemmed from how Muse built its share links: each one included the conversation’s title plus a code, all under one domain. That made the links findable through passive DNS searches, a method for looking up historical domain lookups, even though the intended audience was supposed to be just the person who got the link.

Meta told Ars it is pausing the sharing feature while it reviews the report and considers changes to how the links can be discovered. The report also noted that a separate security flaw in Muse, found weeks after launch, had let other apps on the same device take over a user’s Muse account.

Why it matters

If anyone on your team used Muse’s share feature to send a conversation to a colleague, client or anyone else, that content may have been discoverable by strangers, not just the person you sent it to. For a small business, that could mean loan details, revenue figures, customer information or staff schedules sitting in a publicly findable web page.

Meta has paused sharing while it investigates, but the incident is a reminder that AI assistants with deep access to your accounts and documents carry real exposure risk if their sharing or security features aren’t solid.

What it means for your business

If you run the business: check whether anyone on your team has used Meta’s Muse, especially its share-link feature, for work conversations. Treat anything shared that way as potentially exposed until Meta confirms a fix, and avoid sharing sensitive business details through it in the meantime.

What to do

Act now
  • Ask your team whether anyone has used Muse’s conversation-sharing feature for business matters.
  • Stop sharing sensitive content through Muse until Meta confirms the issue is fixed.
  • If you shared anything sensitive via a Muse link, treat it as potentially public and take appropriate steps, such as alerting affected parties or monitoring accounts.
  • Hold off on connecting Muse to business accounts or sensitive data until its security track record improves.

Act now: worth doing this month.

Jargon in this story

AgentPIIData retention

Sources

  1. 1Meta Muse exposes users’ personal and business secrets when they share conversations (opens in a new tab) · Ars Technica · News report · Oct 20, 2026

Keep going: learn it, then do it.

01Prompter12 min

Know when not to trust the answer

Spot made-up facts, check sources, and learn which questions AI is bad at.

Learn · 5 minTry it · 7 min
Start lesson
Data, privacy & security

What happens to what you paste into a chatbot?

Where your text goes, who can see it, and the settings that matter.

6 min