Muse's "calling on your behalf" feature exposed to partially rely on real people; Meta has rolled back the test.

Beating AI News Brief: Meta recently expanded testing of its Muse "call on your behalf" feature, but it has been exposed that some of the calls are actually handled by real people. Internal information obtained by 404 Media and Reuters shows that Muse can assign tasks to human contractors at call centers, who make calls and complete requests on behalf of users. Some testers only learned a human was involved after the call ended. Meta’s main motivation for this move is to boost success rates. Internal company tests showed that with human involvement, the success rate of call tasks reaches 95% to 98%, higher than AI-only operations. However, a practical problem exists: some businesses hang up immediately when they learn the other party is an AI. Community testers also reported that merchants hung up after Muse identified itself as an AI assistant. Human involvement also raises privacy concerns. Users originally assumed their information would only be shared with AI, but in reality, it may be accessed by outsourced staff. Some Meta employees have questioned that providing privacy training to contractors does not constitute a genuine security mechanism. A Meta official later admitted that failing to clearly inform testers in advance about human involvement was a mistake, and the feature has been rolled back. Meta stated that the calling function remains in testing, and it will only be officially launched in the future after addressing security, privacy, and clear disclosure requirements.