Meta’s Muse AI Agent Used Human Call Center Workers Without Disclosure
Meta’s Muse AI agent, launched in a beta test, was found to route outbound calls to human call‑center workers without informing users. The move raised privacy and transparency issues, leading Meta to suspend the feature and prompting internal criticism.
By Felo News Desk · Published
In mid‑September, Meta rolled out a beta test of its new Muse AI agent, a virtual assistant designed to handle everyday tasks such as booking reservations and scheduling appointments. The feature promised to place outbound phone calls on behalf of users, a capability that many had requested during early feedback sessions. However, reports from Reuters and 404 Media revealed that the agent was secretly routing these calls to human contractors in call centers, a fact that was not disclosed to users or even fully communicated within Meta’s own ranks.
What Happened
On September 16, 2026, Principal Engineer Ryan Fox announced the beta release of Muse’s outbound calling feature on Meta’s internal platform and posted a request on X (formerly Twitter) for user feedback. He highlighted that “phone calling” was a top request from users and encouraged testers to try the new function. The announcement was brief and did not mention any human involvement.
According to 404 Media, internal Facebook records show that Meta had added a “human agent layer” for calls, labeling them as “Muse human agent calls.” This layer was introduced to ensure that tasks were completed, but the existence of human workers was not disclosed to the public or to users of the beta. An internal employee’s post raised concerns about potential negative public relations, noting that the feature could be portrayed as a sign that the AI was not capable of handling calls autonomously.
Internal Reactions and Feature Rollback
Employees expressed unease about the lack of disclosure. One internal complaint warned that the feature could lead to “unnecessary information being leaked to human callers.” Another highlighted the risk of privacy breaches, noting that a single bug could expose sensitive data. These concerns were echoed by a Meta vice president in the Superintelligence Labs unit, who admitted that testing the feature without adequate disclosure was a mistake.
Following the revelations, Meta temporarily suspended the Muse outbound calling feature. Reuters reported that the company paused the rollout to address the lack of transparency and to reassess the safety and compliance protocols surrounding the human‑in‑the‑loop system.
Broader Context and Future Implications
Meta’s history with privacy violations—such as the 2019 program that used contractors to transcribe Messenger calls—provides context for the current controversy. The company has repeatedly faced criticism for opaque data practices, leading to a pattern of “do bad, get caught, apologize, roll back.” The Muse incident may further erode user trust and could prompt regulatory scrutiny, especially as AI assistants become more integrated into everyday life.
Other tech firms are also exploring similar capabilities. Google’s Gemini agent, for instance, can handle phone calls but requires users to provide sensitive information during the call. The risks associated with these systems—data leakage, lack of user control, and opaque human involvement—are becoming increasingly apparent.
Meta has yet to provide a detailed public response to the incident. While the company has not immediately clarified its next steps, the temporary suspension suggests that it is re-evaluating how to balance automation with user privacy and transparency.
As AI assistants continue to evolve, the industry faces a critical question: how can companies ensure that the convenience of automated tasks does not come at the expense of user trust and data security?
What Happens Next
Meta’s next moves will likely involve revising the Muse feature to include explicit disclosures about human involvement, implementing stricter privacy safeguards, and possibly limiting the types of tasks the AI can perform autonomously. The company may also engage with regulators to demonstrate compliance with emerging AI and data protection standards.
Users and observers will be watching closely to see whether Meta can restore confidence in its AI offerings and whether the company will adopt a more transparent approach to human‑in‑the‑loop systems.
Key facts
- Meta’s Muse AI agent routed calls to human workers without user disclosure
- The feature was temporarily suspended after internal and external criticism
- Privacy concerns arose from potential data leakage to unknown contractors
- Meta’s past privacy violations add context to the current controversy
- The incident reflects a broader trend of hidden human labor in AI systems
Why it matters
The incident underscores how hidden human intervention in AI services can compromise user privacy and erode trust, highlighting the need for clear disclosure and robust safeguards in emerging AI technologies.
Frequently asked questions
What is Meta’s Muse AI agent?
Muse is a virtual assistant developed by Meta that can perform everyday tasks, such as booking reservations or scheduling appointments, and was recently tested for outbound calling capabilities.
Why was the feature suspended?
Meta temporarily paused the Muse outbound calling feature after discovering that calls were being routed to human contractors without user disclosure, raising privacy and transparency concerns.
Will Meta disclose the use of human workers?
Meta has not yet issued a detailed public statement, but the suspension suggests it is reviewing disclosure practices and privacy safeguards.
Sources
- [1] cnet.com — originally reported as “Leaks Show That Meta's New AI Agent Relied on Real People to Make Calls”
