Meta Tests Human Backup for Muse AI Assistant Phone Calls

Meta's AI assistant gets human help

Meta has been quietly testing a "human concierge" system for Muse, its new personal AI assistant, where human contractors handle some of the phone calls initiated by the digital agent, according to internal company posts reviewed by Reuters.

The social media giant informed employees about the test last week, shortly after publicly launching Muse's phone calling capability. The arrangement has sparked internal debate about privacy safeguards and the potential for sensitive user information to reach call center workers.

Meta spokesperson Daniel Roberts characterized employee response as "overwhelmingly positive" and framed the test as a feedback mechanism. "We're working with merchants to continue improving this potential calling feature, and will only roll it out when it's ready and with the proper disclosures," Roberts said.

Why it matters

The revelation underscores a persistent challenge in AI development: the gap between what automated systems promise and what they can reliably deliver. When AI assistants make real-world phone calls on behalf of users—to book appointments, make reservations, or handle customer service—the stakes for accuracy and privacy increase dramatically. Human backup systems may be necessary during development, but they introduce the exact privacy risks that AI automation is supposed to eliminate. For enterprise leaders evaluating AI agents, this highlights the importance of understanding what happens when systems fail and whether human oversight creates new vulnerabilities.

Privacy concerns from inside

Some Meta employees have raised flags about the human involvement, warning that contractors handling calls could inadvertently access sensitive personal information. The concerns reflect broader industry tensions around AI transparency—users interacting with what they believe is an automated system may not realize humans are listening or intervening.

The test comes as Meta positions Muse as a more capable successor to its earlier AI assistants. The phone calling feature represents a significant expansion beyond text-based interactions, allowing the AI to make real-world transactions and arrangements on a user's behalf.

The hybrid approach

Using human operators to supplement AI systems during testing and development is not uncommon in the technology industry. Many companies have employed similar strategies to train models, handle edge cases, and ensure quality before full automation. However, the practice becomes controversial when users aren't clearly informed that humans may be involved in what appears to be an AI-only interaction.

Meta's statement emphasized that proper disclosures would accompany any public rollout, suggesting the company recognizes the need for transparency about human involvement. The timeline for a broader release remains unclear, with Meta indicating it will proceed only when safety and privacy protections are fully implemented.

The details were first reported by Reuters based on internal company communications.