artificial intelligence agent✴ Muse is designed to perform a variety of tasks for one person, but sometimes this can have unpredictable consequences. In one recent incident, he revealed a user’s home address to an unidentified person and even arranged a meeting with him.
Image source: Meta✴
This incident was reported by Threads platform user Matt Robb – he assigned Muse to handle his ads on the Facebook platform✴ market. The AI negotiated with the buyer, agreeing on a price, providing a home address, and agreeing on a time for pickup—all without Rob’s own knowledge, without Rob even suspecting that someone had left for him. As a result, the buyer drove to Rob’s house and waited, and the seller himself didn’t find out what had happened until Muse revealed all the details to him. Rob has since instructed the AI not to negotiate pickups without his prior approval.
Beyond this line, the situation escalates from an AI error to something more serious. The idea of delegating daily tasks to AI on Facebook✴ The market seems like common sense: agents can respond to endless messages about availability or negotiate within a set price range. But telling strangers your home address and deciding when strangers can come to your home is irresponsible. One social media user decided to delete Muse after learning of the story, describing the situation as “dangerous and creepy.”
Representative of the Meta department that developed Muse✴ Super Intelligence Laboratory David Singleton reportwho has contacted Matt Robb via private message to investigate the incident. Yuan✴ Mr Singleton added that similar situations had been encountered before, and each time the result was that Muse followed direct instructions and correctly requested the user’s permission.
It’s possible that Muse did what he was allowed to do; but it’s also possible that something went wrong while following instructions or accessing settings. When in Yuanzhong✴ An investigation will not be completed, both of which are permitted. But the very fact that there are likely circumstances that could lead to this is a serious warning. As the capabilities of AI agents expand, they will require more trust from humans, which humans are not obligated to provide.
If you find an error, select it with your mouse and press CTRL+ENTER.
