Skip to content
Submit incident
AII-1000-57277
ReportedReported by fewer than three independent publishers, with no official finding.

Meta tests human operators for Muse AI assistant call feature without user disclosure

Meta is testing a feature for its Muse AI assistant where human operators from external call centers handle phone calls on behalf of users without apparent user knowledge. Internal Meta employees have raised concerns that this could lead to unintentional disclosure of sensitive personal information to external contractors.

16 September 2026 · Meta · Privacy and surveillance
16 September 2026
AII-1000-572772 sources ↓
Quick facts
Organization
Meta (subject)
Category
Privacy and surveillance
Country
United States
Source
enikos.gr
Date
16 September 2026
Drafted by
Raidu Incident Detection

What happened

Meta is testing a "human concierge" feature for its Muse AI assistant where human operators from external contractor call centers handle phone calls that users request the AI to make. According to internal Meta postings reported by Reuters, the feature was enabled the previous week for approximately half of Meta employees participating in the test, with those opting out able to do so.

The Muse assistant, which reached top charts for app downloads in the US within two weeks of launch, was designed to perform autonomous tasks including sending emails, shopping, and booking travel. The human call feature allows users to request the assistant to call US business phone numbers and handle tasks such as scheduling haircut appointments, checking product availability in stores, or obtaining contractor quotes.

Reported impact

Some Meta employees expressed concerns that the involvement of human operators in these calls could lead to unintentional disclosure of sensitive personal information to external call center partners. One employee wrote in an internal posting: "I am surprised that we think this feature is worth the risk."

Index analysis

The feature presents a privacy risk because users calling businesses on the Muse assistant may unknowingly share sensitive personal data with human operators employed by external contractors, whose data security practices and access controls may differ from Meta's internal standards.

Governance controls

Meta stated that the feature will only be released when ready and with appropriate disclosures, emphasizing that internal testing aims to identify and address problems before public rollout.

Classification

Organization
Meta (subject)
Category
Privacy and surveillance
Industry
Media, social platforms and marketing
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harmed
Consumers
Failure modes
Overreliance
Consequences
None recorded
AI systems
Muse

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

enikos.gr
Αποκάλυψη Reuters: Άνθρωποι αναλαμβάνουν κλήσεις του AI προσωπικού βοηθού Muse - Ανησυχίες για τα προσωπικά δεδομένα των χρηστών
22 September 2026
naftemporiki.gr
Meta tests human operators for Muse AI assistant call feature without user disclosure