How much AI is too much?

I started hosting things myself about 5 months ago. It began with classics like Immich and Vaultwarden. The setup went well until I hit my limits when it came to integrating network-related features like Crowdsec or automated offsite backups. That’s when I started using Claude Code, among other things. At first, I just asked Claude for help with these topics and handled the configurations myself. But as is often the case down the rabbit hole, I wanted more: Jellyfin, .arr stacks, etc. Eventually, though, Claude had sudo/root access to my server. I’d just give prompts, and Claude would set things up, troubleshoot, optimize, and automate. It’s great, and it’s allowed me to implement things I wouldn’t have had the know-how to do even after a year. Today, however, I decided to take a step back. I’ve removed root access. scaled back the runbooks, and Claude now only knows the basics. I’m not learning anything myself by using AI. I do enjoy the end product, but at what cost? Of course, I’m also worried about my data security and… who knows where that data might end up?! How do you handle this? How much access do you give AI to your system, and do you worry about data privacy issues? Maybe I’m overthinking when I see how much AI slop is being generated for all sorts of things. I’d love to hear your thoughts :)