Surprise, Meta's latest AI gimmick is just underpaid humans

404 Media has revealed that Meta's latest AI product, Muse — pitched as an AI that makes phone calls for people too anxious to call a barbershop themselves — is actually humans making calls to other humans. Meta chief AI officer Alexander Wang had claimed the previous week that Muse was expanding into phone beta use for making outbound calls to US businesses.

An internal message board post reviewed by 404 shows how Meta obscures the human labor behind the product. According to the post, "Muse doesn't just dial a number. It calls a business on your behalf, handles the conversation, completes your request, and reports back with a transcript and a summary," and "It also no longer works alone. Muse is now able to hand requests to a trained human agent, who places the call and works it through." The practical effect, the piece argues, is that Muse functions as little more than an auto-dialer with transcription capabilities, since a human agent handles the actual communication.

Internally, Meta employees recognized the reputational risk. One warned co-workers that a human-driven AI making calls for users has potential for "so much negative PR," predicting "It could portray us as 'their AI is not good enough so they still need humans' kind of coverage for this launch." Another post pleaded, "Please PLEASE do not release this as-is and at least make this a user preference if we absolutely must launch this. Please." Meanwhile some employees praised Muse publicly on X, where AI boosters amplify hype around the technology.

The report situates Muse in a broader pattern of AI products being exposed as hidden, underpaid human labor, typically performed by workers in countries like Venezuela and Kenya who train and moderate models — and in the US, by people cleaning human feces from the backs of Waymos. Companies want customers to believe the capability is magic, the argument goes, but it is exploitation that solves the problem of calling people.