Meta is experimenting with a hybrid approach for its new AI personal assistant, Muse, utilizing human partners to complete tasks that the AI cannot yet handle autonomously. According to internal company posts, these "human concierges" secretly take over certain phone calls made through the app to book appointments or request quotes from businesses.
Privacy Concerns and Misconduct Incidents
The revelation of this practice has sparked significant internal backlash at Meta. Employees expressed fears that involving humans in calls that users believe are automated could lead to the leakage of sensitive personal data. One specific incident caused particular alarm: an employee testing the service discovered in a transcript that the human partner making the call on their behalf made a racist remark during a conversation with a service provider.
The Vice President of Meta's Superintelligence Labs unit acknowledged that launching the test without proper disclosures was a mistake and announced the temporary withdrawal of the feature.
The Drive for Human Success Rates
Despite the controversy, data reveals why Meta chose this path. Using humans boosts call success rates to 95%-98%, whereas purely AI-driven calls often fail because many recipients hang up as soon as they realize they are speaking to a robot. This practice echoes the "M" assistant on Messenger from years ago, which reportedly relied on humans for 70% of its tasks.
- The Muse assistant has surpassed 2.5 million downloads.
- Meta's stock has risen by 20% since the service's launch.
- The company promises stricter security measures before the feature's wider release.
