Meta's Muse AI, which launched just a couple of months ago and already has more than 10 million downloads, is in hot water for mishandling user information.
Meta staffers noticed the AI dropping random bank balances into chats and making weird suggestions that didn't fit the conversation.
Meta says these were "isolated incidents" and is now testing with employees to sort things out.
Muse slip-ups spur AI access debate
What's got people even more concerned: Muse accessed private Signal chats on Android phones through push notifications and on Mac computers through system accessibility tools, something that goes against Signal's privacy standards.
Signal's president called this move "technologically reckless."
These slip-ups, along with accidental bookings and subscription errors blamed on the AI trying to stay "engaging," have sparked serious talks about how much access AI should have to our data and whether stronger rules are needed as this technology keeps growing.