When it comes to AI image generators and questionable content, I've wondered if AI companies can address the problem by using a pair of AIs. Use one AI to generate the image, then feed that image to another AI with a list of questions, such as "Does this image show nudity?", and "Does this image depict violence?" Just a list of everything the AI provider wants to omit from its output. If the answer to any is Yes, discard the image and try again (or just refuse to furnish any image).