Meta wants its AI to make your calls... even if that means entrusting them to strangers

On September 8, Meta officially rolled out its new model, called Muse, in the United States. It is an agentic AI and is therefore capable of sending emails, booking a trip, and performing all kinds of tasks online. Muse should therefore also soon be able to make phone calls on the user’s behalf.An appointment to make with the doctor or a question to ask a company? Based on a few instructions, this AI can handle it and then provide the user with a report on the call. But the AI could also transfer the call to a human being.

A call delegated to an AI that then transfers it to a call center

Logically enough, Muse’s call feature should be handled entirely by artificial intelligence. The American company has also placed particular emphasis on respecting the privacy of its model. However, according to a Reuters report, which may have had access to internal Meta documents, Muse could also delegate certain calls to contractors, in other words, to humans. This is not merely a rumor but a feature that is being tested. In a Meta document, it states that “Muse is now able to transfer requests to a trained agent, who places the call and follows up.”

A Muse user can therefore delegate a call to the AI, which could in turn transfer it to a human being. Of course, the fact that contractors handle calls placed via the AI agent inevitably raises confidentiality concerns.

What about the confidentiality of these calls? 

According to Reuters, some Meta employees warned that sensitive information could end up in the hands of the call center’s contractors. If the call is simply to make an appointment at the hairdresser, that should not raise too many issues. But what about a medical call or any confidential request made to a company or government agency, to name just a few examples?

unsurprisingly, Mark Zuckerberg’s company expressed confidence in delegating some calls to humans. Meta spokesperson Daniel Roberts told Reuters that employees’ feedback was « extremely positive » and that the test would enable the company to refine its security and privacy-protection measures before the feature’s launch. He added that Meta would deploy it « only once it was ready and with the appropriate information », without specifying what that information would be or who would receive it.

We therefore still do not know the scope of this feature. Will users be able to completely prevent the AI from delegating a call to a human? In any event, this feature undoubtedly has a Kafkaesque dimension.

👉🏻 Follow the tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.

Source: Reuters