Meta has temporarily halted an internal test using human contractors to quietly complete phone calls for its new personal AI assistant 'Muse' following employee backlash over privacy risks and improper conduct.
Launched recently the Muse has gained over 2.5 million downloads by promising to autonomously complete tasks like shopping, booking appointments and placing phone calls on a user’s behalf. However, when businesses hung up on automated AI calls the Meta introduced a "human concierge" feature, assigning call-center contractors to place calls for half of its employees.
Internal company posts seen by Reuters revealed strong pushback from Meta staff. Employees warned that using contractors exposed sensitive personal data to third parties, violating the app’s core security promises. Concerns escalated after a transcript showed a contractor making a racist comment while negotiating a cable bill for an employee.
Following the backlash an executive in Meta’s Superintelligence Labs admitted it was a mistake to run the test without proper disclosures and confirmed the feature had been rolled back. Meta argued that human contractors boosted call success rates to 95–98% compared to lower success rates for AI-only calls.
A Meta spokesperson clarified that the feature was only an internal test meant to gather feedback and refine privacy protocols before any public release. The incident highlights a broader tech industry pattern of relying on hidden human labor to bridge the gaps in autonomous AI capabilities.
