Somewhere in America, a salon receptionist may soon pick up the phone to find a bot on the other end. Meta's consumer AI agent, Muse, can now place phone calls on behalf of paid subscribers in the United States, dialing businesses and carrying conversations from start to finish. The rollout follows an earlier testing phase that Meta pulled back after privacy concerns arose over the involvement of human contractors.

The pitch is refreshingly low-tech in spirit: users tell the assistant what they need, and Muse handles the dialing, the waiting, and the talking. It also reflects a broader shift in consumer AI, where assistants are increasingly expected to complete real-world tasks rather than simply answer questions.

What Muse's Calling Feature Actually Does

The feature allows Muse to make outbound calls to businesses across the US. The use cases are practical: booking appointments and asking for information.

Outbound calling began as a beta in September 2026, and Meta broadened access through a public beta test by mid-September — a compressed timeline that shows how quickly the company is moving agent features from limited testing to general availability.

Calling is one item on a longer list of chores Muse is designed to handle. The agent can also send emails, shop, and book travel on a user's behalf, positioning the app as an end-to-end task runner rather than a question-answering chatbot.

Meta launched Muse in September 2026, and the app found an audience quickly. It climbed to the top of US app download charts and passed 2.5 million downloads within its first few weeks — scale that now matters directly, since the calling feature is reserved for paid subscribers.

The Human Contractor Detour

Earlier versions of the feature involved limited testing in which human contractors stepped in to help with calls whenever the AI ran into trouble.

That setup drew privacy concerns, and Meta rolled it back. For a user, the issue was easy to grasp: you think software is handling your call, but a stranger may be listening in.

Meta acknowledged the misstep. The company indicated it would pursue an improved approach in which the AI handles calls on its own, with the required privacy disclosures in place. How those disclosures are implemented in practice is the detail to watch as calling scales from beta to a subscriber-facing feature.

Muse Voice Brings Avatars Into the Mix

Calling is part of a broader voice push inside the Muse app. On November 5, 2026, Meta rolled out Muse Voice to all US users, adding real-time voice conversations and video chats to the app.

Users can talk with customizable animated avatars that display different expressions and gestures. The feature is limited to users aged 18 and older, who can start calls and chats with the avatars while Muse keeps working on other tasks in the background. Users can also choose from a large selection of voices.

Behind the scenes, the infrastructure numbers are notable: Muse can run 12 concurrent real-time sessions per Nvidia GB200 chip at approximately 870 ms latency. In practical terms, a single chip sustaining a dozen simultaneous live voice sessions with sub-second responses is the kind of per-chip efficiency that makes real-time voice workable at consumer scale.

Hardware Is Next on the Roadmap

Meta does not intend to keep Muse confined to a phone screen. Planned expansions include AI Glasses and a new wearable called the Muse Charm, which is anticipated by the end of 2026 — a near-term deadline worth watching as Meta works to move the assistant beyond the app.

What This Means

Gating calling to paid subscribers ties one of Muse's most tangible features directly to revenue, turning the app's fast start on the download charts into a subscription pitch.

Privacy remains the biggest open question. The contractor rollback showed that Meta's users and critics are paying attention to who—or what—is on the line, and the company has now publicly committed to an AI-only model with the proper disclosures in place.

Source: CryptoBriefing