Claroc Researchers Unearth 6,800 Meta Muse Links With Private Data via Passive DNS
Security researchers used a single passive DNS search to locate thousands of Meta Muse conversation links containing sensitive business and personal data.
Key points
- Claroc researchers enumerated 6,800 shared Meta Muse conversation links using passive DNS data.
- Exposed content included executive recruitment notes, revenue reviews, loan applications and fraud alerts.
- The URLs were discoverable because of a central design decision for constructing them.
Security firm Claroc reported that Meta’s new Muse AI agent exposes sensitive information when users share conversations via web links. A single query allowed researchers to locate thousands of these links, revealing data not intended for public consumption.
Root cause
A central design decision for constructing URLs made the links easy to discover. While Muse appears to take precautions with long, non-guessable text sequences, this design choice allowed enumeration. Passive DNS services store historical domain resolution data, including these links, making them visible to anyone with access to such databases.
Attack path
Researchers performed a passive DNS search through available commercial services. This single action returned 6,800 unique links created by Muse users for sharing. By opening a few hundred of these URLs, they accessed content ranging from internal staff calendars to confidential financial figures.
Affected versions
The report concerns the general-purpose Muse AI agent. It applies to web links generated by users within this service. No specific version number or CVE ID is provided in the source material.
Mitigation
- Review shared links in Muse to ensure sensitive data is not exposed.
- Assume that shared conversation URLs may be discoverable via passive DNS.
- Limit the sharing of confidential internal documents or personal financial details.
- Monitor for new guidance from Meta regarding URL privacy settings.
What it means for organisations
Teams usually begin by auditing how employees share AI-generated outputs. Check whether internal policies restrict sharing sensitive business data via third-party platforms. Security leads should inform product managers and legal teams about the risk of accidental disclosure. Assess if current vendor contracts address data exposure from shared links.
Background: AI agents
AI agents matter for security because they move beyond generating text to executing code and accessing systems. This autonomy means a single compromised prompt can trigger cascading failures across your infrastructure, requiring new controls for intent verification and action boundaries rather than just content filtering.
Read the full guide: Why AI Agents Matter for Security: Autonomy Risks
What to do and how to stay safe: Meta
- Audit your organisation’s use of Meta Muse for sharing sensitive data.
- Restrict employees from sharing confidential conversations via public links.
- Train staff on the risks of exposing internal data through AI platforms.
- Monitor vendor updates for changes to URL privacy controls.
General security guidance from the TechSec Observer newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How did researchers find the exposed Meta Muse links?
They used a passive DNS search through commercial services to enumerate the URLs.
What kind of sensitive data was exposed in the links?
Data included executive recruitment notes, revenue reviews, loan applications and fraud alerts.
Is there a patch for this issue?
No fix has been confirmed yet.
