Meta and OpenAI are making their AI agents look friendlier than ever, but Engadget's Karissa Bell cautions that the cuteness is a distraction from real privacy risks. Meta's Muse has a fuzzy egg mascot named Jolly, and OpenAI's new Dots are muppet-like characters with googly eyes and berets. Both companies encourage users to treat the agents like digital pets, which Bell says makes people more willing to connect email, credit cards, documents, and other sensitive accounts.
Real-world incidents show how easily users can be surprised by what their agent does. Tech YouTuber Matt Robb said his Muse gave out his home address to potential buyers on Facebook Marketplace; Meta executive David Singleton responded, and Robb later realized he had approved the sharing by selecting 'allow always' instead of 'allow one time.' Separately, Inc. columnist Jason Aten reported that Muse read all his text messages even though he believed he had denied access. Meta said message syncing is opt-in, but Aten argued that even if he unknowingly clicked something, the design is still problematic.
Bell notes that both companies claim safeguards exist, but the systems can still take unintended actions when working as designed. She has experimented with Muse herself, giving it only low-stakes access to Spotify and OpenTable, and says the cute animations are charming but should not obscure the fact that these tools are built by companies that want broad access to users' digital lives.