Not really, because Nightshade should have made image labeling more difficult, but if you try it, you'll see that it doesn't do anything; multimodal models are

Not really, because Nightshade should have made image labeling more difficult, but if you try it, you'll see that it doesn't do anything; multimodal models are too powerful nowadays to be fooled by small adversarial noise generated using CLIP LPIPS (small enough not to be too noticeable to us). And Glaze does not try to interfere with labeling.