Meta Muse is testing a “human concierge” system in which contractors handle some phone calls placed through the personal AI agent, as Meta evaluates the feature’s safety and privacy protections before any wider release.
NEW YORK: Meta Muse is testing an unusual approach to artificial intelligence assistance by allowing human contractors to handle some phone calls placed through its new personal AI agent.
Meta informed employees about the “human concierge” experiment last week, shortly after publicly launching phone-calling capabilities for Muse, according to internal company posts reviewed by Reuters. The feature is being tested internally and has not been announced for a general public rollout.
How does the human concierge work?
Muse is designed to perform tasks rather than simply answer questions. Meta says the personal AI agent can send emails, book travel, complete forms and work across online services on a user’s behalf.
Its newer calling capability extends that idea into conversations with businesses. Users can delegate errands that require phone calls, with Muse contacting businesses and providing transcripts and summaries afterwards. During the experiment, however, some of those calls can be handled by human contractors rather than entirely by AI.
Why is Meta testing humans alongside AI?
Meta says the experiment is intended to collect feedback and help the company improve the feature before potentially releasing it more widely.
Company spokesperson Daniel Roberts told Reuters that employee feedback had been “overwhelmingly positive”. Meta said it would introduce the potential calling feature publicly only when it was ready and appropriate disclosures were in place.
The approach highlights one of the challenges facing increasingly autonomous AI agents. Digital assistants may be capable of navigating websites and completing routine tasks, but unpredictable real-world conversations can create situations where human assistance remains useful.
Privacy questions emerge around the experiment
Some Meta employees have raised concerns about the possibility of sensitive information being unintentionally shared with contractors handling calls, according to Reuters.
That issue is particularly relevant because personal AI agents can potentially interact with significant amounts of user information.
Meta says Muse operates through a dedicated Muse Secure VM and that users control which services the agent can access. The company also says credentials are kept in secure storage and that Muse seeks approval before sensitive actions such as purchases or sending emails.
Muse becomes Meta’s push into personal AI agents
Meta launched Muse in the US on September 8, 2026, positioning the product as an AI agent capable of carrying out everyday tasks and longer-term projects.
Meta Muse has since attracted rapid consumer interest, with Reuters reporting more than 2.5 million downloads and a rise to the top of US app charts within two weeks of launch.
The human concierge test shows how companies developing autonomous AI may combine automation with human support while they work to make increasingly complex agent-based services reliable enough for everyday use.
Impact to expect
The experiment could influence how personal AI agents handle complicated real-world interactions. If human assistance becomes part of agent services, technology companies may face greater pressure to clearly disclose when people rather than AI systems are handling user requests, particularly when personal information is involved.
