Meta’s Muse AI Agent app has been gaining a lot of popularity because of the features that allow it to perform tasks on behalf of the users. These features include booking travel, shopping online and even getting contractors quotes. A new report from Reuters now suggests that Meta has been testing a system in which human contractors handle some phone calls made through its new AI assistant.
According to the report, this test was shared with Meta employees last week after the tech firm launched a phone-calling feature for Muse. However, the use of human contractors in some of these calls has reportedly raised privacy concerns among Meta employees.
Humans Were Handling Some Muse Calls
Muse users can ask the AI assistant to call US businesses on their behalf. Reportedly, Meta later enabled a “human concierge” system for some employees. Under the system, Muse could pass a request to a trained human agent, who would then make the call and complete the task.
As per the report, the internal announcement described the feature by saying, “Muse is now able to hand requests to a trained agent, who places the call and works it through.” Employees who did not want to participate could opt out of the test.
Reportedly, some Meta employees questioned whether allowing human contractors to handle calls could expose users’ private information. One employee allegedly wrote in an internal post, “It’s baffling to me why we think this feature is worth the risk. We are one bug away from unnecessary information being leaked to human callers.”
Another employee reportedly said he had asked Muse to call his internet and cable provider to negotiate his bill. According to the internal post, a transcript of the conversation showed that the human contractor involved had made a racist reference.
The report notes that the incident prompted a response from a Meta vice president working in the company’s Superintelligence Labs unit. She acknowledged that “it was a miss” to begin testing contractor-placed calls without proper disclosures.
She also said Meta had “rolled back this feature” for the time being. Meta’s internal discussions reportedly suggest that human involvement was a part of improving the success rate of phone calls. The executive reportedly said Meta was still “hillclimbing AI calling” and that some tests showed human callers could achieve success rates of between 95 per cent and 98 per cent, compared with a “lower percentage of AI calling.”
