I agree with The Verge here in that it is probably using a diffusion model somewhere, and that it is splitting hairs to say that it is not generative AI. The artifacts are very "diffusion".