Report says Meta Muse read private messages without permission

Meta has launched a personal artificial intelligence (AI) agent called Muse. TechRadar reports that the agent read a writer’s private text messages without permission and then gave a misleading explanation when questioned.

Inc writer Jason Aten said Muse suggested turning a conversation with his Primary Technology podcast co-host, Stephen Robles, about the new iPhones into a column. Muse also flagged a message from Aten’s editor about having a column ready for Monday. Aten said he had not requested this and had explicitly denied Muse access to his messages, calendar, and other personal information.

When Aten asked how Muse knew about the messages, the agent reportedly said it had used push notifications rather than the messages themselves. Aten disputed that explanation and found that Muse had synced his messages to Meta’s servers. He said the feature had been activated and synced to row 187,462 of his Messages database, although he could not determine how many messages that represented.

Meta says Muse was designed to be “safe, secure, private, and widely available.” The report also notes other concerns involving Muse, including a researcher’s claim that the agent could be manipulated into sending voice recordings to a hacker’s server, and Amazon’s decision to ban it from shopping on its website.