AI copyright news, graded by source

News publishers ask the court to deny OpenAI and Microsoft summary judgment, arguing training and RAG on their articles are not fair use and that no implied license exists

In a 76-page omnibus brief filed Oct. 23, 2026 (public version with redactions; the full brief is under seal at Dkt. 2119), the News plaintiffs oppose OpenAI's and Microsoft's motions for summary judgment. They argue that copying articles to train models and to ground chatbot answers (RAG) shares the original purpose of delivering news to readers, so it is not transformative; that under Warhol the defendants lacked justification because their own experts admit the models did not need News content to generalize; that the articles were copied whole; and that the defendants, not the publishers, bear the burden on market harm. They reject an implied license, saying copyright is not an "opt out" regime and that they objected through terms of use, bot blocking and these suits. They argue OpenAI's content extractors were designed to strip copyright management information, and that statutory damages run per article. These are the plaintiffs' arguments; the motions are undecided. OpenAI's opposition papers to the authors' own motion were filed the same day, largely under seal.