From the article:
>In SafetyKit’s evaluations involving drawings or animations that sexualize children, the classifier detected 90% of cases
>This is the first CSAM classifier with broad availability that has a specific signal on indicating if CSAM may be AI generated.
What children are being abused by drawings or AI generated images, stories, etc?
I understand the cases where it's problematic to make such content based on real children and share it, and I think it's probably a bad idea to allow photorealistic child porn to be publicly posted on the internet for the sake of not confusing law enforcement (in the same way posting misinformation is harmful), but if these images are not realistic, or if it's just text, or if they are realistic but they aren't being shared on the internet, who is being harmed?