# Meta Pauses Muse Sharing Feature After Security Researchers Expose Public Chat Links Publicly discoverable URLs led to the exposure of sensitive corporate and personal user data. - **Source:** [The Verge (theverge.com)](https://www.theverge.com) - **Published:** 2026-10-20 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/e1266bea-873d-467d-bc1f-b2aaf1693bdd ## Feature Suspension Following Discovery Meta has officially paused the Muse feature allowing users to share AI conversation links. This decision follows a report from security firm Claroc, which identified over 6,800 publicly accessible links. The company is currently reviewing its sharing infrastructure to prevent further unauthorized access to private user interactions and sensitive data. ## Predictable URL Generation Mechanism Claroc researchers discovered that Muse constructed web addresses using a combination of the conversation title and a random ID. Because hostnames often revealed content before access, the predictable structure allowed researchers to reconstruct links via public DNS records. This design flaw effectively bypassed intended privacy controls for shared AI artifacts. ## Exposure of Sensitive Information The investigation revealed that exposed chats contained highly confidential materials. Found data included a marketing firm’s revenue reports, small business loan applications, and internal workplace calendars. Additionally, notes regarding job candidate interviews with an AI lab were accessible, highlighting the significant risks associated with sharing AI-generated content via public web links. ## DNS Records as Vulnerability Public DNS records, which function as directories for web traffic, served as the primary vector for this data exposure. By querying these records, Claroc was able to index hundreds of private conversations. This incident demonstrates how standard web infrastructure can inadvertently expose private AI interactions if URL generation lacks sufficient entropy. ## Industry-Wide Privacy Challenges This vulnerability is not unique to Meta, as similar issues have plagued other major AI platforms. Conversations from OpenAI’s ChatGPT, Anthropic’s Claude, and xAI’s Grok have previously appeared in public Google Search results. These recurring incidents underscore a systemic challenge in balancing user-friendly sharing features with robust data privacy and security protocols. ## Meta’s Official Response Muse spokesperson Tracy Clayton confirmed the pause in an emailed statement to The Verge. The company is currently evaluating how these links are discovered and indexed. Meta aims to implement necessary changes to the sharing functionality before restoring it, ensuring that user conversations remain private and protected from unauthorized public access. ## Operational Impact on Users The immediate suspension of the sharing feature disrupts workflows for users who relied on Muse for collaborative tasks. While the pause is temporary, it forces a re-evaluation of how AI-generated content is distributed within professional environments. Users must now wait for Meta to provide a more secure method for sharing their AI interactions. ## Future Security Implications The Muse incident serves as a critical reminder for developers regarding the security of AI sharing features. Moving forward, platforms must prioritize non-deterministic URL structures and stricter access controls. Protecting user data in the age of generative AI requires moving beyond simple link-based sharing to more secure, authenticated methods of content distribution. ## Sources covering this story - https://www.theverge.com/tech/1014627/meta-muse-ai-sharing-ai-chats — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._