Meta Tests Muse AI Agent Calls That Are Actually Made By Humans in a Call Center
404 Media Jason Koebler ● Covered by 11 sources
Meta said Muse could make calls for you. It turns out humans in call centers may be making some of them.
Based on reporting by 404 Media, Jason Koebler — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Meta has been pitching Muse as an AI agent that can phone a business, book a table, and set up appointments on your behalf. But according to internal material seen by 404 Media, the company has also been testing a human fallback: if the system needs help, a trained person can take over and place the call.
That is a big difference from the public framing. On Meta’s internal message board, the company told employees it had “added a human agent layer for calls to get completed” and said “Muse human agent calls” were ready for dogfooding. The same post described Muse as more than a dialer, saying it handles the conversation, completes the request, and returns a transcript and summary. It also said the system can hand work to a human agent.
The problem is obvious. Users may think they are speaking to software when a person is actually hearing their request, and that can get awkward fast if the call involves something personal, like a doctor’s appointment. Employees on the inside seemed to see the risk too. One worried about the privacy implications of passing user requests to other human beings; another said the feature could trigger headlines about Meta’s AI not being good enough to work alone.
Meta’s own messaging to staff tried to calm those fears, arguing that contractors had received a lot of training and that data was safe and secure. An employee pushed back on that logic, saying training is not a security mechanism. Some testers said they were not told a human had made the call until after it happened, which only sharpens the concern.
The company says this is still confidential and pre-launch, and that it will only roll out the feature with proper disclosures. That may be the only sensible answer here. If an AI feature quietly depends on humans to finish the job, users deserve to know before they hand over anything sensitive.
My take — AI-written commentary, not fact-checked reporting
This is the usual AI theater: sell autonomy, quietly staff the seams with people, then call it innovation. Meta is not alone, but it does seem unusually eager to make “AI” sound like magic while hoping nobody notices the call center underneath. That trick is getting tired, and users are the ones paying for the smoke machine.
Read more about this at: 404 Media
