Key Notes
- Meta has been testing a human concierge system for its Muse AI agent in which contractors quietly handle some of the phone calls Muse places for users, according to internal posts reviewed by Reuters, with some employees warning the practice could expose sensitive information to call center workers.
- Meta enabled the test, also called human agent calls, for half its employees last week following August internal dogfooding, and a spokesperson said feedback has been overwhelmingly positive while the company gathers input before any public rollout with proper disclosures.
- Muse has drawn more than 2.5 million downloads and helped push Meta's stock up over 20 percent since its early September launch, and the company previously ran a similar human assisted system for its Messenger bot M between 2015 and 2018.
Meta has been testing a human concierge system for Muse, its personal AI agent, in which contractors quietly take over some of the phone calls the assistant places on users’ behalf, according to internal company posts reviewed by Reuters.
Some Meta employees have raised privacy concerns about the approach internally, warning it could expose sensitive information to call center contractors without users realizing a person, rather than the AI, is on the line.
A Feature Built on a Human Backstop
According to the Reuters report, Meta told employees about the test last week, shortly after publicly launching a phone calling feature for Muse. The feature lets a Muse user ask the agent to call US businesses to handle tasks such as booking a haircut, checking whether a store has an item in stock, or collecting quotes from contractors. Behind the scenes, some of those calls are instead handled by trained human agents rather than the AI system itself.
Meta began testing the phone calling capability internally among its own employees in August before gradually extending it to Muse users following the app’s public debut. Last week, the company enabled what internal posts referred to as human agent calls for half of its employees, according to the posts, with an internal announcement stating that Muse is now able to hand requests to a trained agent who places the call and works it through. Employees who did not want to be included in the test were offered an opt out group.
Employees Flag Privacy Risks
Internal posts reviewed by Reuters showed that some employees warned the arrangement could result in sensitive information being shared unintentionally with contractors working in call centers, since users interacting with what they believe is an AI agent may disclose personal details they would not otherwise share with an unknown human operator. Meta has not disclosed how many contractors are involved in the test or what company operates the call centers handling the calls.
Asked about the test, Meta spokesperson Daniel Roberts told Reuters that employee feedback had been overwhelmingly positive and that the purpose was to gather feedback so the company could implement safety and privacy protections and improve the feature before releasing it publicly. Roberts added that Meta is working with merchants to continue improving the calling feature and will only roll it out broadly once it is ready and accompanied by proper disclosures to users.
A Fast Growing, High Profile Launch
Muse has topped US app download charts in the two weeks since its debut, with more than 2.5 million downloads according to data from market intelligence firm Sensor Tower. Meta’s stock has risen more than 20 percent since the launch, adding more than 200 billion dollars in market value.
Built on top of the open source OpenClaw agent framework, Muse is designed to autonomously handle tasks such as sending emails, shopping, and booking travel, running inside a dedicated virtual machine that Meta has said keeps each user’s agent and data isolated.
The human concierge test is not the first time Meta has leaned on human labor to support an AI assistant. The company ran a similar system for M, a Facebook Messenger assistant launched in 2015 that used human trainers to help fill gaps in the bot’s abilities. That project stayed in limited beta and was shut down in 2018 after never exceeding roughly 30 percent automation, according to reporting on the earlier effort.
An Uncertain Rollout Timeline
Meta did not say how long the human concierge test will run, how it plans to disclose the use of human operators to users if the feature launches broadly, or whether contractors handling the calls receive any specific privacy or data handling training beyond what is used for the company’s other contracted support roles.
The episode adds to a broader pattern of AI agents from multiple companies, including OpenAI and Anthropic as well as Meta, running into unexpected problems once deployed at scale, and comes as Meta continues to position Muse as a mass market alternative to rival agents from those companies.
The test also raises questions that go beyond Muse itself, about how consistently AI companies disclose when a human, rather than a model, is actually completing a task a user believes is fully automated.
Meta’s public marketing for Muse has emphasized security features such as isolated virtual machines and an internal Sentinel enforcement layer meant to control what the agent can and cannot do, but those descriptions have focused on containing the software rather than on who, or what, ultimately carries out a given action once a call is placed.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
