MENLO PARK, CALIFORNIA / RankWire.AI / – In the United States, on October 20, Meta has put a hold on its Muse AI assistant’s chat sharing feature after security researchers demonstrated that thousands of shared chat links could be found through public internet records. This exposure potentially risked revealing sensitive personal and corporate data. Meta announced on Tuesday, October 20, that it had halted the feature pending further review of the report. The company is considering modifications to how shared links are accessible.
The discovery originated from security company Claroc, which identified over 6,800 Muse share links within passive DNS records. These records log Domain Name System lookups, making website URLs accessible to anyone searching the data. Claroc stated it notified Meta through its bug bounty initiative on October 8, and made its findings public on October 18. The 6,800 figure reflects the number of discovered addresses, not sensitive chat content.
Researchers from Claroc examined several hundred of these conversations, and nearly all live shares could be accessed without requiring a login. Included among the visible data were a small-business owner’s loan application that contained part of an account number and a note about a bank fraud alert. Other exposed information included a marketing firm’s revenue review prepared for its CEO and a real-estate investor’s report detailing mortgage information and returns. Confidential interview preparation notes were also visible in the shared chats.
Hostnames revealed chat titles
The core issue related to the design of Muse’s sharing URLs, which incorporated both the conversation title and a random identifier within the hostname. While the identifier was not easily guessable, visiting these links triggered DNS lookups that passive DNS services could log. This meant that individuals who never received the link directly could nonetheless find it stored in these databases. These shared conversations could be accessed without any login or sign-in process.
Although Muse’s settings instructed search engines and AI crawlers to ignore its share pages, researchers did not rely on search engines alone to locate them. Over 75 saved scans of Muse share pages were stored on urlscan.io. Claroc reported that there was no evidence of data misuse. The firm did not contact users affected by the leak and could not determine whether others had discovered the links. Meta’s statement did not clarify if users would be notified about the exposure.
Feature temporarily halted while Meta reviews findings
On October 20, when Claroc re-examined 40 previously accessible shares, none loaded, and their addresses returned errors. Despite this, the conversation titles still appear in passive DNS history. Muse was launched in the United States on September 8 for adults aged 18 and over. It offers integrations with users’ email, calendar, payment, and health applications. In April, Meta introduced its Muse Spark model, developed by its superintelligence lab.
This incident underscores ongoing concerns over privacy in AI applications and the dangers associated with sharing chats containing sensitive information. In 2025, conversations shared via ChatGPT were found in Google search results, leading OpenAI to remove the indexing feature. Additionally, in August, Meta was ordered to fund teen mental health programs in New Mexico. Meta has not disclosed when Muse’s sharing functionality will be restored or how the link security measures will be modified.