The fact that OpenAI already has public moderation APIs for detecting prompt safety, I'd guess there is a lot more internal work going on in determining what pr

The fact that OpenAI already has public moderation APIs for detecting prompt safety, I'd guess there is a lot more internal work going on in determining what prompts to ignore, or what phrases of one's prompt to adjust to avoid responses like this.