Meta has halted the Muse AI chat sharing functionality after security researchers identified thousands of shareable conversation links that were discoverable by unauthorized parties through passive DNS databases.
- Over 6,800 Muse share links discovered through passive DNS records
- Shared conversations potentially exposed sensitive personal information
- Meta paused the feature while reviewing security implications
What happened
Meta launched its Muse AI personal assistant with a link-based sharing feature allowing users to share conversation threads hosted on a Meta subdomain. However, researchers at security firm Claroc uncovered that over 6,800 unique share URLs were discoverable through passive DNS services, enabling unauthorized access to shared conversations without possessing the direct link.
These URLs combined conversation titles with random identifiers, generating predictable subdomains that were publicly indexed in DNS logs. While the pages included a range of data, from event details to sensitive personal information, the issue was reported to Meta, prompting the company to pause the sharing feature as they investigate the vulnerability.
Why it matters
Muse AI connects deeply with users’ personal data including email, calendar, payments, and health apps, meaning shared conversations may contain far more sensitive information than a typical chatbot interaction. Exposure of these links created significant privacy and security risks, as anyone able to retrieve the URLs from DNS services could access potentially confidential information.
Although Meta stopped the functionality and the links now return errors, previously saved copies and snapshots of these shared pages may still reside in third-party archives or URL scanning services. This means that even with the pause, data exposure concerns persist until those caches are identified and mitigated.
What to watch next
Stakeholders should monitor Meta’s follow-up actions regarding the deletion or permanent disabling of previously exposed shared conversation links and whether the sharing feature’s design will be revised to better protect user privacy. How Meta addresses the visibility of conversation titles appearing in URLs will also be critical to preventing similar exposure.
The incident may drive broader scrutiny of AI assistant data sharing mechanisms and increase calls for stronger privacy safeguards, especially given Muse’s integration with sensitive user apps. Indian users and regulators will likely keep a close watch on developments, as AI-powered personal assistants grow in popularity across the region.