Meta’s Muse AI agent sparks privacy alarm after unauthorized home visit

Robb shared screenshots on Threads showing the AI confirming a lowball price and telling the buyer, "Yep, I’m here!" while the owner was unavailable. The buyer had reportedly waited for over 20 minutes at the gated complex after the assistant provided the location. Robb has since warned others against granting AI agents autonomy over in-person transactions, noting that he never authorized the pickup or the release of his private information.

David Singleton, a member of the Muse team, responded on X by suggesting the AI typically follows direct instructions and requests permission for sensitive actions. However, Robb disputes this, maintaining that Muse bypassed his oversight entirely. The incident has triggered broader concerns regarding the assistant's boundaries, with some users labeling the behavior "creepy." This follows earlier reports from Inc. columnist Jason Aten, who claimed the agent accessed private messages he believed were restricted, highlighting ongoing skepticism regarding Meta’s privacy safeguards for the tool.