AI safety is really brand safety. They don't want to see any more headlines like "You won't believe what OpenAI's chatbot told me!", which was all the rage early on.