Meta told employees that when its Muse agent places calls for users, a trained human in a call center dials, talks and finishes the job.
Last week Meta executives announced on X that Muse, its AI assistant, could call businesses for you to book a table or a haircut. The internal notice the same day said something else: the company had added a layer of real people to get the calls completed.
AI on the poster, a row of agents at the desks
The internal post is blunt. Muse hands requests to a trained human, who dials the number, handles the conversation, finishes the request and reports back with a transcript and a summary. It also tells employees not to share any of it, because the product has not launched.
Staff did not understand it. 'Why is this a feature?' one asked on the internal board. Another predicted that if it shipped on by default, the coverage would write itself: Meta puts humans behind the scenes, and all the goodwill from early adopters evaporates.
A colleague who tested it said they only learned a person had made the call after it ended, having assumed the AI was talking.
An old trick, and the same old risk
Tech companies passing humans off as AI is a long tradition. The dangerous part is the information. Ask Muse to book a specialist appointment and it may not occur to you that a contractor just read it. Meta's answer is that these workers get extensive training; an employee replied that training is not a security mechanism.
Meta told 404 Media that internal dogfooding is part of development, that employee feedback has been overwhelmingly positive, and that the feature will only ship when ready and with proper disclosures.
Why it matters
Booking a table, a clinic slot, a doctor's appointment — these are things people are careful about even saying aloud. If you think you are talking to a machine and a person is listening, the privacy line has not been blurred. It has been quietly removed.
Curated from high-quality sources, with concise summaries and key takeaways.
