There's more CSAM per capita (at least 10x likely more like 100x) on a platforms which "purports to police CSAM" like Civit.ai. No one does anything about it though.
Oh you don't believe me? This is one of the most popular models on their whole site: https://civitai.com/models/402667/age-slider-lora-or-ponyxl-...
Stable Diffusion 1.5 was trained on at least thousands of images of this stuff and no one has tried to get that model taken down, mostly because training on sufficiently large datasets almost guarantees that you'll have illegal content included within it. https://www.theverge.com/2023/12/20/24009418/generative-ai-i...
I'm also shocked for related reasons that the 3 and 4 letters haven't been more upset about LLMs, since there's almost certainly a ton of classified info just sitting on the internet that LLMs have ingested into their training data. (don't believe me? https://www.airforcetimes.com/news/your-air-force/2024/03/26...)
