The United States tells a court that training is fair use
The Department of Justice filed a statement of interest in the consolidated litigation against OpenAI: copying protected text to train a language model is fair use.
Why it matters
For the first time the federal government stated a position on whether training on protected works infringes copyright, in the case that will set the American rule.
The statement was filed on 1 September 2026, runs to 19 pages and was made under section 517 of title 28 of the United States Code, in the consolidated proceeding In re OpenAI Copyright Infringement Litigation in the Southern District of New York, before Judge Stein. The government states that the United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally. The copying of protected text articles as part of training a language model is described as a use of a different kind or character that is transformative, spectacularly so. The government calls the Kadrey court's application of the fourth fair-use factor, with its broad market dilution theory, deeply flawed, arguing that outputs merely existing in the same genre as training materials are not cognisable market harm, and that training is its own use, to be analysed separately from output uses. The filing carries no binding force: it is a way for the United States to state a position without becoming a party.