Meta Rolls Back Human-Assisted AI Calls After Racist Incident and Privacy Concerns

Meta’s Muse AI platform, launched this month, quickly became one of the most downloaded apps on U.S. iOS devices. However, its rapid rise has been overshadowed by reports of human contractors stepping in to assist AI agents during calls—including an alleged incident involving racist remarks and subsequent privacy concerns. The company has since rolled back the feature.

Launch and Early Success

Meta’s Muse AI platform, launched on September 8, quickly gained traction, topping the U.S. iOS download chart. The platform includes a beta feature allowing AI to make phone calls on behalf of users, positioning it as a tool for tasks like scheduling appointments or checking service availability.

Human Contractors in the Loop

According to Gizmodo, Meta tested a “human concierge” feature within Muse, where contractors could intervene when the AI struggled with a task. This hybrid approach was designed to handle errands like booking appointments, verifying store inventory, or gathering quotes from service providers.

The feature was part of internal testing, or "dogfooding," a process Meta described as core to its product development. Employees involved in testing were reportedly given the option to opt out.

Incident and Backlash

The hybrid model faced scrutiny after a Meta employee alleged that a contractor made a racist remark during a test call. According to Gizmodo, a vice president at Meta Superintelligence Labs apologized to the employee, and the contractor was subsequently dismissed.

The incident raised broader concerns among Meta employees about the ethical implications of involving human contractors in AI-driven interactions. Some worried that the feature could inadvertently expose sensitive user information to third-party operators.

Feature Rollback

Meta rolled back the human contractor calling feature following the incident and internal pushback. Despite tests showing that human-assisted calls achieved success rates in the high 90% range—compared to lower rates for AI-only interactions—the company appears to have prioritized addressing privacy and ethical concerns.

The decision reflects broader challenges in AI development, where balancing automation with human oversight remains a contentious issue. For example, one user reported that insurance companies repeatedly hung up when the Muse AI disclosed it was not human, highlighting another layer of complexity in AI-driven communication.

What this means

LazyFounders analysis — our interpretation, not reported fact.

Meta’s challenges with Muse highlight the tightrope AI platforms walk between automation and human oversight. While blending AI with human support can improve performance—especially in complex or sensitive interactions—it also introduces risks like privacy violations, ethical lapses, and user distrust.

For founders building AI-driven products, this episode underscores the importance of rigorous safeguards before deploying hybrid systems. It’s not just about technical success rates; it’s about ensuring transparency, accountability, and alignment with user expectations. Meta’s swift rollback of the feature suggests that even tech giants can misjudge the balance between utility and risk, a lesson worth heeding for startups scaling AI tools.

Key takeaways

• Meta’s Muse AI platform gained rapid traction but faced backlash over its hybrid AI-human testing model.
• Human contractors were involved in calls to assist AI, raising ethical and privacy concerns.
• A contractor’s alleged racist remark during a test call led to an apology and termination, highlighting risks in outsourcing human oversight.
• Meta rolled back the feature after internal pushback, despite high success rates in human-assisted calls.
• The incident reveals the complexities of scaling AI tools that rely on human intervention, particularly in high-stakes interactions.

FAQ

What is Meta’s Muse AI platform?

Muse is an AI-driven platform launched by Meta that includes a beta feature allowing AI to make phone calls on behalf of users for tasks like scheduling appointments or checking service availability.

Why did Meta roll back its human-assisted calling feature?

Meta rolled back the feature after an alleged incident where a contractor made a racist remark during a test call. Internal concerns about privacy and ethics also contributed to the decision.

How successful were the human-assisted calls compared to AI-only calls?

According to reports, human-assisted calls achieved success rates in the high 90% range, while AI-only calls had lower success rates.

What risks did Meta’s hybrid AI-human model pose?

The model raised concerns about privacy, as sensitive user information could be exposed to human contractors. Ethical issues, such as misconduct by contractors, also emerged as a significant risk.

Related on LazyFounders

• Meta
• More ai news
• Meta tests hybrid human-AI system for Muse AI agent