Meta pauses Muse sharing feature after researchers uncover 6,800 exposed user links
Meta is temporarily suspending the conversation-sharing feature in its new AI agent, Muse, after security firm Claroc identified a vulnerability allowing enumeration of 6,800 user-generated links. The flaw exposed sensitive business and personal data, including executive search notes and marketing strategies. This development challenges CEO Mark Zuckerberg’s recent claim that Muse was built for privacy, prompting Meta to review its URL construction logic to prevent unintended public access to user data.
Key points
- Meta, the parent company of Facebook and Instagram, is suspending the sharing feature in its new general-purpose AI agent, Muse, to address a privacy vulnerability.
- Researchers from security firm Claroc enumerated 6,800 unique links created by Muse users to share conversations, revealing that the URL structure allowed easy discovery of sensitive content.
- Exposed data included confidential executive-search notes naming internal contacts at frontier AI labs and strategic marketing documents from national companies.
- Meta CEO Mark Zuckerberg previously described Muse as being built from the ground up for privacy and security, a claim now under scrutiny due to this design oversight.
- Meta is currently reviewing the feature and considering changes to the URL generation process to ensure that shared conversations remain accessible only to intended recipients.
- The incident highlights the security risks inherent in AI agent interfaces, where standard sharing mechanisms may inadvertently expose proprietary business intelligence to the public web.
Meta has temporarily suspended the conversation-sharing feature in its new AI agent, Muse, following a security review that revealed significant privacy gaps. The decision comes after researchers from the security firm Claroc demonstrated that users’ shared conversations could be easily discovered by third parties. The vulnerability stems from the way Muse constructs unique URLs for sharing content, a design choice that made it possible to enumerate thousands of links without needing the specific URL string.
Claroc’s team identified 6,800 links created by Muse users for sharing searches, conversations, and images. Upon analyzing a sample of these links, the researchers found data that was clearly intended to be private. Notable examples included a candidate’s preparation notes for a confidential executive search call with a leading AI lab, which revealed the name of an internal contact and the meeting time. Additionally, a national marketing company’s strategic documents were exposed, offering insights into their internal operations. These findings suggest that the standard security measure of using long, unguessable URL sequences was not sufficient in this specific implementation.
The incident poses a challenge to Meta’s recent marketing of Muse. Last month, CEO Mark Zuckerberg stated that the platform was built from the ground up for privacy and security. However, this oversight indicates that the underlying architecture may have prioritized ease of sharing over robust access control. Meta is now reviewing the feature and considering modifications to how URLs are generated to prevent unintended public access. For global users and businesses adopting AI agents, this event underscores the need for careful vetting of data-sharing mechanisms in new software tools.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.
