MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has paused the sharing of conversations within its Muse AI assistant following discoveries by cybersecurity experts that thousands of shared chat URLs could be located via public internet records. These links potentially revealed sensitive personal and corporate data. On Tuesday, October 20, Meta informed Ars Technica that the feature has been temporarily disabled while it investigates the findings. The company is also exploring modifications to how shared links are accessed and found.
The security firm Claroc identified over 6,800 shared Muse links through passive DNS records, which log domain lookups and allow web addresses to be searched. This figure accounts for discovered URLs rather than the total number of sensitive conversations. Claroc published its findings on October 18, stating it initially reported the issue to Meta’s bug bounty program on October 8, providing reproduction steps and example shares.
A number of these conversations were accessed by researchers, with nearly all of the live shared links opening without requiring a Meta account. Some of the chats contained highly sensitive information, such as an internal revenue review for a marketing firm and a small-business loan application. Others included a bank fraud alert with a transaction amount and partial account details. Additionally, confidential interview notes for executive searches and a workplace vacation calendar listing dozens of employees were discovered. Many of the other shared chats contained typical, routine content.
Weak link design made chats discoverable
The vulnerability originated from how Muse generated its share links. Each URL combined the conversation’s title with a randomly created identifier embedded within the hostname. Although the identifier was sufficiently lengthy to prevent guessing, opening a link triggered a DNS lookup, which passive DNS services could record and index. This enabled others to locate the URLs without direct access and view the chats without needing to sign in.
Although Muse requested search engines and AI crawlers not to index its shared pages, this did not prevent access via DNS records. The security site urlscan.io stored more than 75 saved scans of Muse share links. Claroc noted it found no evidence of misuse but also stated it did not contact affected users or determine if others had already accessed the links. Meta’s official statement did not clarify whether affected users would be notified.
Sharing of links has now been disabled following review
On October 20, Claroc re-examined 40 shared links that previously opened without an account, and none of them loaded. The URLs serving the content returned errors, but conversation titles still remained in passive DNS history. Meta launched Muse in the United States on September 8, targeting users aged 18 and older. The assistant is capable of connecting to users’ email, calendar, payment, and health applications. In April, Meta introduced its Muse Spark model developed within its superintelligence lab.
This incident underscores the growing concerns surrounding data privacy in AI applications and highlights the dangers of sharing sensitive information through chat links. Similar issues surfaced in 2025 when shared ChatGPT conversations appeared in Google search results. Meta has also faced legal challenges related to user safety, including a court order in August requiring the company to pay $567 million. The company has not announced when Muse’s sharing feature will be restored.