Exclusive-Meta testing a 'human concierge' for its new personal AI agent Muse

Meta is testing a "human concierge" component for its new personal AI assistant, Muse, according to internal company documents seen by Reuters. The test involves having human contractors handle some of the phone calls initiated via Muse, following the public launch of a phone-calling feature for the digital agent. The Muse AI, which has been among the top US app downloads since its debut, is designed to autonomously perform tasks such as sending emails, shopping, and booking travel on behalf of users.

However, some employees have raised concerns about privacy, warning that the approach could lead to sensitive information being unintentionally shared with contractors. In response, Meta's Superintelligence Labs unit acknowledged the mistake and rolled back the feature temporarily. The company is still fine-tuning the AI calling capability, with tests suggesting that human assistance could improve success rates from the low to mid-90s percent range.

Meta has faced criticism for its emphasis on safety protections, with employees warning that relying on human labor behind the scenes could undermine the AI's key selling points.