Meta Muse share links can be enumerated, exposing conversation privacy
Understand the story
On October 20, 2026, security firm Claroc released a report stating that Meta's general-purpose AI agent Muse has a privacy leak risk in its conversation sharing feature. The root of the problem is that the share link's URL contains the conversation title and uniformly uses the cf.metaaiusercontent.com domain, making links easy to enumerate. Through a passive DNS query, security researchers discovered 6,800 Muse share links; after opening some of them, they saw sensitive content including executive headhunting call notes, loan applications, bank fraud alerts containing partial account information, and internal corporate revenue reviews. The report reveals the scale and sensitivity of the leak; the incident is still developing, with no follow-up reports of an official Meta response or fix.
AI overview of reporting · Updated 3 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Ars Technica · AIMeta Muse exposes users' personal and business secrets when they share conversations
Security firm Claroc reports that share links for Meta's general AI agent Muse can be easily enumerated because the URL contains the conversation title and always uses the cf.metaaiusercontent.com domain. Through a single passive DNS query, researchers found 6,800 shared links; opening some revealed sensitive content such as executive headhunting call notes, loan applications, bank fraud alerts containing partial account numbers, and internal revenue reviews.
Story heat trend
Current heat9·Comparable peak10（10/20 22:00）·Comparable change over 24 hours –
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.