AI Safety Idea: Whenever an LLM roleplays or generates fiction, its text box should display a semi-transparent visual watermark/overlay. This instantly signals non-factual content, preventing confusion and jailbreaks. #LLMs #AISafety #Transparency