MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has temporarily disabled the chat sharing feature within its Muse AI assistant after security researchers demonstrated that thousands of shared links could be found through public internet records. This vulnerability potentially exposed sensitive personal and corporate data. On Tuesday, October 20, Meta announced it had halted the feature while it examines the incident. The company is considering modifications to how shared links are accessible.
The discovery originated from security firm Claroc, which identified over 6,800 Muse share links within passive DNS logs. These logs record Domain Name System lookups, making website addresses available to anyone inspecting the data. Claroc reported that it alerted Meta via its bug bounty program on October 8 and published its findings on October 18. The 6,800 figure reflects the number of addresses found, not sensitive chat content.
Researchers at Claroc examined several hundred of these conversations, discovering that nearly all shared links were accessible without requiring login credentials. Among the exposed data were a small-business owner’s loan application containing part of an account number, and a note regarding a bank fraud alert. Additional examples included a revenue report prepared by a marketing firm for its CEO and a mortgage and returns analysis from a real-estate investor. Confidential interview notes were also visible in some cases.
Hostnames exposed chat titles
The core issue stemmed from how Muse’s sharing URLs were structured. Each link incorporated the conversation title and a random identifier within the hostname. Although this identifier could not be easily guessed, visiting the link triggered a DNS lookup that passive DNS services could log. As a result, individuals who never received the link could discover it through these databases. Importantly, the shared conversations could be accessed without signing in.
While Muse’s configuration instructed search engines and AI crawlers to ignore its share pages, researchers found that they did not need a search engine to locate them. Over 75 snapshots of Muse share pages had been stored on urlscan.io. Claroc clarified that there was no evidence of data misuse and did not reach out to affected users, noting it was uncertain whether others had already found the links. Meta’s statement did not specify whether users would be notified about this issue.
Sharing feature suspended during review of security report
When Claroc revisited 40 previously accessible shares on October 20, none of the links loaded successfully, displaying errors instead. Nevertheless, the conversation titles remained visible in passive DNS records. Muse was launched in the United States on September 8, targeting users aged 18 and older. The app can connect to users’ email, calendar, payment, and health apps. In April, Meta also introduced its Muse Spark model, developed by its superintelligence laboratory.
This incident heightens ongoing concerns regarding privacy and security in AI tools, especially when sharing conversations that may contain sensitive information. In 2025, shared ChatGPT interactions appeared in Google search results, leading OpenAI to remove the feature that allowed indexing. Additionally, in August, Meta was ordered to fund teen mental health initiatives in New Mexico. The company has yet to specify when Muse’s sharing functionality will be restored or how it plans to modify link accessibility.