So quick #asknostr

You have an encrypted group chat app (like slack / discord), you want to add an AI model to do things for you. 

But it can't read anything because everythings encrypted. 

Giving it general access to groups is a big no no becuase it basically leaks everything to the server and the AI provider!

Best UX I can come up for for this, is 

1) Give the AI assistant an NSec and treat it like a user.
2) As it has an NSec the AI can decrypt any messges sent to it in a DM or group. 
3) You then treat it "like a human" UI wise and it has to be explicitly added to a group before it can read anything? 

Anything obvious I'm missing here with this?