When they discover teenagers smoke and drink - which also exposes them to ridicule, whether they do it or not - it's going to shock them. If federal law, physical ID verification, active law enforcement, massive fines to parents doesn't eliminate drinking and smoking, why would people expect simple age limits stop signing up online?

And yes, online activities open up an avenue for bullying. So does pretty much everything. We should work on that aspect, of preparing children for it and also cracking down on it.

Perhaps a good step for AI chat bots would be monitoring online chat services and learning to detect bullying and taking actions: block all communication, warn a guardian, etc. Detecting harassment could be achievable long before real intelligence is achieved. And detection lets the child, the parent, or the service take many actions in response.