The problem with AI is that combining things is not difficult. For example, if I prompted an AI to generate an image of a cat riding a velociraptor while holding a vacuum cleaner it will be able to generate it. The model knows what a cat is, what a velociraptor is, and what a vacuum cleaner is. It likely has not been trained on many images of a cat riding a velociraptor whie holding a vacuum cleaner.
An AI will know what a child looks like and some AI models have been trained on adult sexual images. What has to happen to prevent the two from being combined is that the harness controlling the LLM has to be tightly programmed to reject the prompt. I think that's the crux of the lawsuit here, that Grok is not effectively blocking the prompts from being run and is not doing an adequate job of banning users who attempt such things.
An AI will know what a child looks like and some AI models have been trained on adult sexual images. What has to happen to prevent the two from being combined is that the harness controlling the LLM has to be tightly programmed to reject the prompt. I think that's the crux of the lawsuit here, that Grok is not effectively blocking the prompts from being run and is not doing an adequate job of banning users who attempt such things.
