Meta is testing a human support layer for its new personal AI Agent Muse as the company works to expand the agent’s ability to handle real-world tasks. Under the test, human contractors can step in and handle some phone calls that Muse would otherwise manage on its own.
The move highlights both the potential and the challenges of building AI agents that can act independently on behalf of users. While Muse is designed to automate everyday tasks, the involvement of human contractors has already raised questions about privacy and how personal information is handled.
Muse Moves Beyond Traditional AI Assistants
AI Agent Muse is designed as an autonomous personal assistant rather than a chatbot that only responds to questions. It can perform tasks such as sending emails, shopping, booking travel and interacting with businesses.
Meta launched the AI Agent Muse in the United States earlier this month, positioning it as an important part of CEO Mark Zuckerberg’s broader vision for personal AI.
The service has gained rapid early attention. Muse recorded more than 2.5 million downloads within two weeks of its launch and reached the top of US app download charts, according to data from Sensor Tower.
AI Agent Muse Can Now Make Phone Calls
One of Muse’s newer capabilities allows it to make phone calls to US businesses on behalf of users.
A person can ask Muse to perform tasks such as booking a haircut, checking whether a product is available at a store or obtaining quotes from contractors. The agent then calls the business and interacts with the person on the other end.
Afterward, Muse provides the user with a transcript and summary of the conversation.
The feature represents a significant shift in how AI assistants operate. Instead of simply providing information, the system can take action in the physical world through phone-based interactions.
Human Contractors Enter the Process
Meta is now testing what it calls a “human concierge” system. Under the arrangement, Muse can transfer certain requests to a trained human agent who makes the phone call and handles the conversation.
The company enabled the feature for about half of its employees during the internal test. Employees who did not want to participate were given an option to opt out.
The testing began after Meta had already started experimenting with Muse’s phone-calling capabilities internally in August.
The human layer appears to be designed as a fallback while Meta continues developing the technology and evaluating how well autonomous calling works in real-world situations.
Privacy Concerns Surface Among Employees
The involvement of human contractors has created a new privacy question.
Some Meta employees expressed concerns that sensitive information could unintentionally be exposed to contractors working in call centers. A phone conversation conducted on behalf of a user could potentially involve personal details, depending on the task.
This creates a different privacy challenge from conventional AI interactions. Users may assume that an AI agent is handling a task entirely through software, while a human may actually become involved in the process.
Meta says the current experiment is intended to help the company understand these issues before making the feature available more broadly. The company has also said any wider rollout would include appropriate disclosures.
Meta Emphasizes Safety and Privacy
Meta has placed considerable emphasis on security around AI Agent Muse .
The company says each user receives a dedicated virtual machine in the cloud where the agent operates. Sensitive information such as credentials is stored separately, while additional controls govern how Muse interacts with external services.
Meta has also built a system called Sentinel that controls whether Muse can perform certain actions through connected services. Depending on the action, the system can require direct user approval before allowing it to proceed.
The company is also developing a confidential virtual machine system intended to prevent even Meta from accessing information stored in a user’s environment.
Human Oversight Could Become Part of AI Agents
The Muse experiment points to an important development in the AI industry. As AI systems become capable of performing more complicated tasks, companies may increasingly combine autonomous software with human intervention.
This approach can provide a fallback when an AI encounters a situation it cannot reliably handle. However, it also introduces questions about transparency, privacy and user expectations.
The central challenge is not simply whether an AI can make a phone call. It is whether users understand exactly who or what is acting on their behalf.
Meta Wants Muse to Become a Personal Digital Operator
Meta’s broader ambition for Muse extends far beyond answering questions.
The company wants the assistant to operate across email, shopping, travel, connected services and other parts of a user’s digital life. Its architecture is designed to allow the agent to work in the background and use tools to complete tasks.
That vision depends heavily on trust. Users must be comfortable giving an AI access to information and allowing it to take actions that can affect their money, communications and daily activities.
The Next Challenge Is Trust
AI Agent Muse rapid adoption suggests that consumers are interested in AI systems that can do more than generate text or answer questions. But autonomous action creates a higher standard for safety and transparency.
Meta’s human concierge experiment shows that even advanced AI agents may sometimes need people behind the scenes while the technology develops.
For Meta, the next stage will be finding the right balance between automation and human oversight. The success of Muse may ultimately depend not only on what the AI can accomplish, but also on how clearly users understand what happens when they hand a task over to it.
Also Read :- Wheat Buyers Brace for Higher Costs as Russia-Ukraine War Disrupts Black Sea Supplies
