Meta Pauses Muse AI Chat Sharing After Researchers Find Public Chats

Meta has paused a sharing feature for its Muse AI assistant after researchers at security firm Claroc found thousands of conversation links publicly accessible on the web. The feature allowed users to share Muse conversations through unique URLs. Claroc identified more than 6,800 Muse links using public DNS records, which act as a directory for web traffic. The exposure stems from how the shareable links were structured. Each URL was built from the shared conversation’s title, a short phrase describing the chat, and a random identifier. In many cases, the hostname revealed the content of the page before opening, allowing DNS records to reconstruct working links. Claroc said it was able to view hundreds of Muse conversations, including materials such as a marketing company’s revenue review, a small business owner’s loan application, a workplace staff calendar, and notes prepared for a job candidate’s call with an AI lab. Muse spokesperson Tracy Clayton said Meta has paused the ability to share specific Muse artifacts while it reviews the findings and considers changes to how these links can be discovered. The pause affects the function that creates accessible pages for people with the link. The incident adds to a broader pattern of AI tools unintentionally exposing user content through shareable links. Similar issues have affected chat interfaces from OpenAI, Anthropic, and xAI, where shared conversations became discoverable through search engines.