Meta pauses Muse chat sharing after researchers find 6,800 public conversation links

What happened

Meta has temporarily switched off a Muse feature that let users share their conversations with the AI assistant through a link. According to The Verge, which cites an earlier Ars Technica report, the company took the step after researchers at the security firm Claroc showed that thousands of those share links were discoverable by anyone browsing public web infrastructure.

Muse spokesperson Tracy Clayton confirmed the move in a statement to The Verge: "People have the ability to share specific Muse artifacts, which creates a page that can be accessed by those with the link. However, we've paused this functionality while we review this report and consider changes to how these links can be discovered."

How the links were found

In a report published this week, Claroc said it uncovered more than 6,800 Muse share links by searching public DNS records, the directory systems that route web traffic to its destinations. The researchers did not need to break into anything; the links were effectively exposed by the ordinary mechanics of the web.

The root of the problem, according to Claroc, was how Muse constructed its share addresses. Each URL combined the conversation's title — a short phrase describing the chat — with a random identifier. Because the hostname carried that same title and ID, a hostname recovered from DNS data was enough to rebuild the full link. In many cases, Claroc said, the hostname alone revealed what a page contained before the researchers even opened it.

What the conversations contained

Working through the recoverable links let Claroc read hundreds of real Muse conversations. Among them, The Verge reports, were a marketing company's revenue review, a small business owner's loan application, a workplace's full-year staff calendar, and notes for a job candidate's call with an AI lab — material that plausibly includes commercially sensitive and personal information.

A familiar failure mode

This is not the first time an AI sharing feature has made private-feeling chats more public than intended. The Verge notes that conversations with OpenAI's ChatGPT, Anthropic's Claude, and xAI's Grok have previously turned up in Google Search results through similar shared-link mechanisms.

Why it matters

The incident is a reminder that "anyone with the link" can quietly become "anyone who knows where to look." Share links are a standard convenience across consumer software, but small design choices — here, embedding human-readable titles into hostnames — determined whether sharing stayed limited or became effectively public.

It also highlights the gap between how people use AI assistants and how that data is handled. Users pasted revenue figures, loan paperwork, and HR schedules into chats, apparently assuming the exchanges were private. Until Meta explains how it plans to change link discovery, anyone who previously shared a Muse conversation has little way of knowing who may have read it. For the wider industry, it is a nudge toward opaque, non-guessable identifiers and sharing flows that treat link exposure as a real threat rather than an edge case.