Meta's new AI is accused of reading all your messages, even if you refuse to allow it.

On September 8, Meta rolled out its new model, called Meta Muse, in the United States. It is capable of answering all kinds of questions, but also interacting with applications and performing tasks on the user’s behalf, such as booking a hotel online.

Meta particularly emphasizes its new AI’s respect for privacy, but reality seems somewhat different.Recently, we learned that Meta Muse could route some phone calls to a call center. Today, a tech journalist reports that Meta’s new AI accessed his iMessages even though he had prohibited it from doing so.

Despite being denied access, Meta Muse uses personal messages to answer a request

Everything begins when Jason Aten, a columnist and tech journalist for the American media outlet Inc., decides to install the Meta Muse model on his iPhone and Mac in order to test it. At first, he asks the AI to write a short biography about him and help him with his work, particularly by suggesting people to interview and topics to cover. So far, everything is going normally. But at some point, Meta Muse suggests that an ongoing conversation between Jason and his podcast co-host about the new iPhones could make for an article. This notification also refers to another conversation with his editor-in-chief about preparing a column.

Most likely, Meta Muse therefore had access to Jason Aten’s personal conversations. The issue here is not that the AI refers to messages, but that it does so even though the journalist had forbidden it access during setup.Indeed, Jason Aten states: “ I remember explicitly choosing not to allow it to access my messages, my calendar, and my other personal information”.

Meta Muse categorically denies having access to the journalist’s message history

The journalist asked Meta Muse whether it had accessed his message history despite the ban, which the artificial intelligence formally denied. The latter explained that it had only had access to the text displayed on the notification banners. Not very convinced by this explanation, Jason Aten conducted his own investigation. He claims to have found that Muse was synchronizing data from the local database of the Messages app on his Mac. And the journalist makes no secret of his dissatisfaction: “Prying through the private database of my messages on my Mac does not reassure me at all. Lying about it is not better”.

This episode therefore reminds us that AI tends to be particularly intrusive, even when we forbid it from prying into certain matters.

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

Source: Inc.