Is anyone surprised that humans are reviewing chats with LLM AI systems?