Fuck AI

Love your point. I wonder if any author or publisher has been held liable for a motivating a murder or a suicide? That would be the legal precedent in my totally ianal mind. Otherwise, I'd think the LLM owner could argue that they're protected by free speech or some other precedent where written word led to harm but was found inculpable.

Because then you'd have to find a judge who would come to the conclusion that the LLM is more than just dynamically written speech. That it was intended to be a treatment for medical purposes and not just "for entertainment". Or that it's effects, as you describe, are more akin to a drug than to reading a book. And while I think future research will absolutely find a link between these chatbots and psychological outcomes, you'd have to find a pretty tech savvy and enlightened judge to make that case based on what we know today. Especially if they're invested in an AI ETF.