AI for kids should be boring - in a good way

The safest AI app for a child is one with no strangers, no ads and nothing leaving the phone. Stella makes bedtime stories on-device for that reason.

What makes an AI app unsafe for kids

Most of the risk rides on the network: ads served to small screens, chats routed through servers, profiles quietly built from a child's name and interests, and open-ended conversation with a cloud model that was never designed with kids in mind. Take the network away and most of those risks go with it. That is the short version of why on-device matters for children.

On-device is the safest setting

Stella generates personalized bedtime stories right on the phone. No account, no feed, no chat with strangers - the AI's entire job is telling a calm story, and nothing your child says or hears leaves the device. Safety here is not a parental control bolted on afterwards; it is the architecture itself. A story generator cannot leak a conversation that never happened.

Personal without a profile

Stories are woven from a name, favorite animals and tonight's theme - personalization that lives inside the app rather than in a database. There are no ads to tap, nothing to buy mid-story, and no internet requirement for anything to leak through. The details you type in stay on the phone they were typed into - no server ever learns what your child likes.

A parent-shaped speed

It does one thing well: a fresh story read aloud with gentle narration and ambient sound, built to wind the evening down rather than crank it up. For a child's first contact with AI, boring and bounded is exactly the right feature set - novelty every night, limits everywhere else. The phone hands over a story, not an open door.