If this takes off, it will be bypassed within a month. Adversarial training is something Stable Diffusion users already invented, and we use it to make our artwork better by poisoning the dataset to teach the network what a wrong result looks like. They reinvented our wheel.