>People have killed themselves or others due to conversations they had with LLMs, but those are just the extreme cases. Most schizophrenics don't commit suicide or kill others, they are mentally ill nonetheless.

Do you think the kind of person who was already psychologically unhinged enough to kill themselves or another person because a chatbot told them to would be completely harmless and totally safe if only chatbots had never been invented? Or is it possible that close to all of the risk posed by this person comes from the person's mental illness, and not the pixels on the screen they're looking at?