The thing about LLMs encouraging people to kill themselves is that as we've seen with all other attempts to implement "safety measures" and "limitations" on these models, they do not work and will not work here either. They will always do shit like this as long as they exist
