The paper’s second corresponding author, and one of the most-cited legal scholars anywhere, Mark Lemley, Law Professor at Stanford, who has represented Meta and Stable Diffusion in model training lawsuits at various points, is very much associated with the view that model training is fair use as long as the model outputs are not infringing.4 He is on record saying that OpenAI will probably have to settle with the New York Times because the Times has found outputs that are copies.