Must an AI be useful? Isn’t being creepy sufficient testament to its utility?
“Creepy” is the word being used to describe Meta’s new AI assistant, Muse. Chances are you’ve seen it pop up on Instagram or Facebook, where Meta is heavily promoting its tool as an AI agent for casual users who are blissfully unaware of what an “OpenClaw” or “Instinct” is.
To use it, you install it on your device and link it to stuff like your email and bank accounts. Then you set it on a task and it “keeps working while you get on with your day,” Meta says. As you keep interacting with the AI, it builds a “curated long-term memory” so it can make more suggestions and better automate tasks.
Here’s the real shocker, though: it’s kind of a privacy nightmare. Many who’ve experimented with Muse were unsettled by how much it dug up on them and how closely it monitored their habits.
Jason Aten at Inc Magazine wrote about how Muse began bringing up specific details he shared in text messages he’d sent through Apple’s Messages app, despite never giving the AI permission to view them. When confronted, the AI explained that it was watching his notifications to glean his message history. It apologized for this transgression, but wouldn’t — or couldn’t — give him a straight answer on how it’d doing it.
Meta Superintelligence Labs employee David Singleton responded to Aten to clarify that the AI wasn’t actually scanning his notifications, though his explanation for what was going on wasn’t any more reassuring. Singleton noted that Muse, if the user gives it permission, syncs data from Messages and asks for full disk access to find files, and implies that Aten must’ve had these permissions enabled. But Aten insists that he never gave the AI any of the required permissions, and says he hasn’t heard back from Meta on an explanation on how this happened.
Not creepy enough for you? Reece Rogers writing in Wired detailed the escalating series of personalized suggestions that Muse began giving while he spent some time with the tech.
“The day after I said I wanted help saving money for a fancy vacation,” he wrote, “Muse offered to ‘wire your savings tracker to your real balance’ if I just tapped a few buttons and linked the data from my checking and savings accounts.”
That wasn’t all. Muse suggested letting it scan his entire email inbox, take pictures of important documents, photograph his meals to track their calories, and share when his passport and license expired.
“Every suggested interaction with Muse started to feel like a guise for me to upload more data about myself for Meta to see,” Rogers opined.
It’s a level of invasiveness that’s quite bold, even for Meta. AI companies usually walk on eggshells so their tech doesn’t come off as too nosy. And then here’s Meta’s AI asking you to fork over your driver’s license and banking info.
The post Meta’s New Muse AI Agent May Be the Creepiest Ever Released appeared first on Futurism.
