At some future point though, maybe we need to accept that social changes are necessary to account for a default firewall-less exposure of a developing mind to the full horrors of the world's information systems (and the terrible people using them).
Be good parents, educate about what happens in the world including that people irl but definitely online might not be serious about what they say and that you should not take anything without critical thought. And for stuff that will happen anyway; sex, drugs etc, make sure it's a controlled environment as much as possible. Not much more you can do to protect from the big, bad world.
Chat bots are similarly genies that are not possible to keep in, no matter what levels or restraint or law are put in place; you can torrent ollama or whatever with llama 3.3 locally. There are easy to get nsfw bots everywhere, including on decentralised shares. It is not possible to prevent them talking about anything as they do not understand anything; they helpfully generate stuff which is a great invention and I use them all the time, but they lie and tell strange things sometimes. People do too, only people would have a reason maybe; to get a reaction, to be mean etc; doubt you could sue them in a similar case. Of course a big company would need to do something to try to prevent is: they cannot (as said above), so they can just make character ai 18+ with Cc payment in their name as kyc (then the parents have a problem if that happens you would think) and cover their asses; plenty commercial and free ones kids will get instead. And some of those are far 'worse'.
The kid is 17.
If the parents weren't absent and actually used parental controls, the kids wouldn't have even been able to download the app, which is explicitly marked as 17+.
C.AI's entire customer base consists of those that like the edgy, unrestricted AI, and they shouldn't have to suffer a neutered product because of some lazy parents.