

This is one of those things where both are likely to be true. All webscale datasets have a problem with porn and csam, and it’s like that people wanting to generate csam use their own fine tuned models.
Here’s an example story. https://cyber.fsi.stanford.edu/news/investigation-finds-ai-image-generation-models-trained-child-abuse and it’s very likely that this was the tip of the iceberg, and there’s more csam still in these datasets.
That’s one of the reasons it’s so good. Alien is perfect space horror. Aliens doesn’t try to match it but tells a different kind of story with some of the same characters.