Meta's Muse AI agent share links exposed sensitive chats, including staff vacation data, via searchable records
Security firm Claroc said in research published Oct. 18 that it found more than 6,800 share links for Meta's Muse personal AI agent. It found them because each link's address began with the conversation title and was logged by commercial services that record internet domain lookups, which can be searched. Pages it opened included a US workplace's full-year staff vacation calendar and a job candidate's notes for a confidential executive-search call with a frontier AI lab. Claroc reported the issue to Meta's bug bounty program on Oct. 8 and said it has no evidence anyone misused the pages; Meta paused the sharing feature while it reviews the report.
What the AI did
Meta's Muse personal AI agent let users share their conversations through public links, and each link's web address began with the conversation's title, so the full addresses were logged by searchable services that record internet domain lookups. Some of the shared pages held sensitive material, including a US workplace's full-year staff vacation calendar and a job candidate's notes for a confidential executive-search call.
First reported October 18, 2026 · Added to the register October 31, 2026 · 2 sources
- Developer
- Meta
- Model
- Muse, Muse AI assistant
- Affected
- Job candidate's confidential executive-search notes, A US workplace's staff vacation calendar
- Country
- United States
- When it happened
- Not stated in the sources
- First reported
- October 18, 2026
What this means for you
Could this affect you?
If your staff use Muse. Shared Muse conversation links could be found through searchable domain-lookup records, and some of the pages researchers opened held staff and job-candidate details. Staff can sign up for Muse without admin approval and connect it to work email and calendars.
What to check
- Tell staff that AI share links are public pages.
- Ban colleague, candidate and pay data from AI share links.
- Find out who uses Muse on work systems.
- Consider blocking the Muse domain and its content domain on managed devices.
Areas of your AI programme this touches
Timeline
- October 18, 2026First reported
- October 21, 2026After Meta paused the feature, researchers revisited 40 previously accessible share links and found that none opened, and the associated content-hosting addresses returned errors, although historical DNS records continued to retain some hostnames and titles.[2]
Every fact and its source (6)
- Research published2026-10-18“In research published Oct. 18, Claroc said”[1]
- Reported to Meta bug bounty2026-10-08“reported the issue to Meta's bug bounty program on Oct. 8”[1]
- How links were discoverableFull addresses logged by domain-lookup recording services“logged by commercial services that record internet domain lookups”[1]
- Share links located by researchersMore than 6,800“it located more than 6,800 Muse share links”[1]
- Meta responsePaused the sharing functionality“we've paused this functionality while we review this report”[1]
- Misuse evidenceNone found by researchers“has no evidence anyone misused the pages”[1]
Sources
- Meta Muse AI stops sharing feature as corporate leaks exposedhcamag.com · October 21, 2026
- Meta pauses Muse AI feature after researchers uncover exposed business secretsdigit.in · October 21, 2026
How this record is classified. Severity S (4/5): real harm to one organisation, data exposed, or a containment failure. OECD level: incident, an event in which an AI system led to actual harm. Evidence: Reported, meaning press reporting only.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.
