Meta Muse AI Assistant: When “AI” Calls, Is It Really AI?

23 September 2026 | By the Supra Vista Team

Imagine asking an AI assistant to call a shop, book an appointment or negotiate a bill.

Now imagine discovering that a human contractor may actually have made the call.

That is what Meta has been testing with Muse, its new personal AI assistant.

Why did Meta use humans?

Muse is designed to do more than answer questions. It can carry out tasks such as sending emails, booking travel and making phone calls on a user’s behalf.

But real-world conversations are unpredictable. People interrupt, ask unexpected questions or may be uncomfortable speaking with AI.

Reuters reported that Meta tested a “human concierge” system in which contractors quietly handled some Muse calls. Internal testing suggested human-assisted calls could achieve success rates of around 95–98%, higher than fully AI-driven calls.

The bigger question: who is really doing the work?

The experiment raised privacy and transparency concerns inside Meta. A Meta executive acknowledged that testing contractor-assisted calls without proper disclosure “was a miss”, and the feature was rolled back for now. Meta says it is continuing to develop AI calling.

Human involvement in AI is not unusual. People help train systems, review outputs and step in when technology fails.

The more interesting question is when users should be told that a human is involved.

If an AI assistant says, “I’ll handle it,” should you know whether the task is being completed by:

AI? A human? Or both?

For families, this is a useful reminder that understanding AI is not only about what it can do — but also how it does it.

Explore our AI courses for children →