How Do You Balance Safety & Engagement in AI for Kids?

A few weeks ago, I made my nephew a personalized Super Mario Brothers chatbot for his 9th birthday using Claude Code. Since he's only 9, I put a ton of security around it and only included info his parents approved - like his first name, sports he plays, and favorite topics.

He loved it initially and would rush to use it after homework, which felt amazing. But he said it just says the same stuff over and over again. I think I made it too strict trying to keep him safe.

Now I'm trying to find the right balance. I left it open so he can explore learning topics, so maybe I'll suggest he ask Mario for help with math since he's struggling with that in third grade. He could really use a companion to work through these things with.

Has anyone else built something like this for a kid? How do you find the perfect balance between safety and keeping it engaging? I'd rather err on the side of too strict, but I want him to actually enjoy using it.

4 comments