Technology and AI
The United States tells a court that training a model on copyrighted text is fair use
By Staff Writer | 4 September 2026
The Department of Justice has filed a statement of interest in the consolidated copyright litigation against a model developer, arguing that training large language models on written works is a use of a different character from the original and that a contrary ruling would threaten national security. The newspaper leading the case says the government is siding with trillion dollar companies against the people whose work was taken.
The statement was filed on 1 September 2026 in In re OpenAI, Inc. Copyright Infringement Litigation, the multidistrict proceeding numbered 25-md-3143 before Judge Sidney H. Stein in the United States District Court for the Southern District of New York. It runs to twenty pages and is filed under 28 U.S.C. section 517, which lets the department attend to the interests of the United States in any pending federal suit without leave and without becoming a party. It is signed by Stanley E. Woodward Jr., the Associate Attorney General, and Brett Shumate, the Assistant Attorney General for the Civil Division.
The lead case is the one brought by a New York newspaper company against Microsoft Corporation and the OpenAI entities, filed on 27 December 2023. The statement says for simplicity that it refers only to that newspaper and that developer, but that its arguments apply to every party in the consolidated cases, including the book authors and publishers.
What the government argues
The interest section comes first. The statement says the United States has a strong interest in continuing to develop a competitive artificial intelligence industry that sets the global standard, cites the President's executive orders of 23 January 2025 and 2 June 2026, and says that effective integration of the technology will be critical to national security capabilities. Rules of law that make it much harder to build the industry domestically, it says, threaten national security and hand an advantage to foreign adversaries who are not so encumbered.
For all these reasons, the United States has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law.
Stanley E. Woodward Jr., Associate Attorney General, and Brett Shumate, Assistant Attorney General, Civil Division, in the Statement of Interest of the United States
On the merits the statement addresses two of the four statutory fair use factors. On the first, the purpose and character of the use, it says copying articles as part of training a model is a use of a different kind from the original, with a further purpose and a different character, and that the difference is exceedingly large. On the fourth, the effect on the potential market for the work, it says a copy used in training does not serve as a substitute for the original and that outputs lacking substantial similarity cannot cause cognisable market harm. It describes a market dilution theory adopted in a Northern District of California decision as contrary dicta that misapplies copyright principles, and says that at a minimum any market effect cannot overcome the value of the training use's new purpose.
The statement also makes an economic argument. If licensing were required, it says, only the largest technology companies could afford the fees, and the fees would disproportionately benefit legacy media because of the volume of their output. It takes no position on whether a licensing regime would be feasible, and notes that publishers can and do license real time and paywalled content to developers whatever the answer on training.
The newspaper's answer
The plaintiff's response came the next day, through a named spokesman.
The Administration is siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole.
Graham James, spokesman for the newspaper company that is the lead plaintiff
He added that both artificial intelligence and creators can thrive if developers pay fairly for the content that makes their products possible, as copyright law requires, and that letting companies take content without permission or compensation would undermine the sustainability of human created content on which the technology itself depends.
A statement of interest binds nobody. It tells the judge where the executive branch stands, in the run of copyright suits brought by authors, publishers and news organisations against model developers.
The docket has moved since. The developer filed its answers to the amended complaints on 3 September 2026. On 31 August the judge ordered the plaintiff to show cause by 11 September why the action should not be stayed pending summary judgment in the other active cases in the multidistrict litigation, with any response from the defendants due by 18 September. The government's twenty pages will be read alongside whatever is filed then.