Meta has suspended the internal testing of a human concierge feature for its Muse AI assistant. The program allowed trained contractors to intervene in outbound phone calls when the AI was unable to complete a task independently.

The trial, which began in mid-September 2026, involved approximately 50% of Meta's workforce. While the company reported positive internal feedback, it decided to pause the initiative to address concerns regarding potential sensitive data leakage.

The Muse AI agent, which launched on September 8, 2026, on iOS, Android, and WhatsApp, has reached over 2.5 million downloads in the United States. The company is currently reviewing its privacy protections before proceeding with further rollouts.