Meta Tests Muse AI Agent Calls That Are Actually Made By Humans in a Call Center

Meta, the social media giant, recently unveiled a new feature for its AI agent, Muse, which allows it to make outbound calls to businesses on behalf of users. However, according to internal sources, Meta is testing the deployment of human agents to handle these calls in call centers. This revelation has raised concerns among employees regarding privacy implications and the perception of Muse's capabilities.

During internal testing, it was discovered that Muse does not always operate solely on AI, but rather hands off requests to trained human agents who complete the calls. This means that users may not be aware that a human is making the call on their behalf. The internal message board post, seen by 404 Media, states that this feature is still confidential and for pre-launch testing only.

Meta's chief AI officer, Alexandr Wang, announced the expansion of the Muse beta for outbound US business calls on September 16, while the company's principal engineer on Muse, Ryan Fox, posted the same information on X. Despite the company's assurances that data shared with human contractors is secure, employees have voiced concerns about potential privacy breaches, especially when sensitive information like doctor's appointments is involved.

One employee expressed skepticism, stating that the feature raises questions about whether Muse is truly advanced enough to handle calls independently. Another employee worried about negative press depicting Meta as relying on human agents when its AI is not sufficient. Meta's spokesperson responded by stating that the company is committed to improving the feature and will only roll it out after addressing safety and privacy concerns.