Meta tests humans secretly taking over some calls made by its Muse AI agent

VraserX · x · 2026-09-23

Meta is reportedly testing human contractors quietly taking over some phone calls made by its new Muse AI agent.

Commenting on the news, VraserX argues the human fallback itself isn't the problem — hiding it is: if you ask your personal AI to call your doctor, bank, or lawyer, you need to know when a random human enters the loop.

His takeaway: human in the loop is fine; secret human in the loop absolutely isn't — raising broader questions about transparency and disclosure obligations for AI agents.

Related event: 404 Media Exposes Meta's Muse AI Calls as Human Call Center Work(7 posts)→

Original post →

More from AGI Musings

AGI Musings channel →