And the list concept concerns me. Are these lists appealable? If not, then they're abusable.
Also, the line between "AI generated" and "non-AI generated" is ever more fuzzy. AI is used for upscaling. AI is used in cameras to enhance images taken. AI is used to make the sort of minor edits that are done the world over in Photoshop. Etc. There's also the fact that this is done with image fingerprinting, which is fuzzy, so then any images that have minor modifications done with AI which get added to the list will get the raw images flagged as well. The thing people want to stop is "fake images", and in particular, "fake images that mislead about the topic at hand". But then that's not "AI" that's the problem in specific, that's image fakery in general (AI just makes it faster / easier).
And re: fingerprinting, take for example, the famous case of the content-spam creator who took a photo of a woodcarving of a German Shepard, flipped it horizontally, ran it through an AI engine to make trivial tweaks to the image, and then listed it as his own. I'd think any decent fingerprinting software would catch both versions. And if it's not flexible enough to catch that, then I have to wonder how useful it is at all, since images constantly change as they move around the internet, even accidentally, let alone deliberately.