On September 1, 2026, the U.S. Department of Justice filed a Statement of Interest in the consolidated copyright litigation against OpenAI and Microsoft that is pending before Judge Sidney Stein in the Southern District of New York. This multi-district litigation case combines the claims of the New York Times and other news-publishers with an Authors Guild class action.
The government’s position is that training a large language model on copyrightable text is most often fair use.
This filing is not a customary amicus brief. The DOJ exercised a right to participate based on 28 U.S.C. § 517, a statute that allows the federal government to appear in any court proceeding to represent the interests of the United States. The Statement of interest claims a US security interest in the case’s outcome: “Rules of law that make it significantly more difficult to develop a robust AI industry in the United States … threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered.”
The Fair Use Argument
Those who have been following the 50+ ongoing cases on AI and copyright infringement will not be surprised at the DOJ’s argument. The brief separates the act of training AI models from the possibly infringing nature of their output in using the models–as have many of the cases on this topic. The Statement of Interest outlines the four statutory fair use factors and lands heavily on the side of the AI developers for training:
Transformative Purpose (factor one, usually). This has typically been the most important factor in recent fair use analysis. The DOJ describes training as transformative (citing Bartz v. Anthropic PBC, 787 F. Supp. 3d 1007, 1021 (N.D. Cal. 2025), which described it as “transformative–spectacularly so.”). This is because the model does not use the copied text for the purpose for which the text was created (informing or entertaining a reader). An AI model converts the text into statistical representations to learn patterns of language. The brief cites Google v. Oracle and the Second Circuit’s Authors Guild v. Google (the Google Books case) for the proposition that copying in service of a new technological function–search, indexing, or here, model training–can be transformative even in cases of extensive copying.
Market Effect (factor four). Plaintiffs in these cases have argued a “market dilution” theory: that even if no single output reproduces a protected work, a model trained on an author’s entire body of work can generate stylistically similar content that competes with, and dilutes the market for, that author’s work generally. DOJ calls this theory “deeply flawed” and specifically the criticizes Kadrey v. Meta Platforms‘ fourth-factor analysis for accepting a version of it. The DOJ’s argument is that training, by itself, does not make any protected expression available to the public, so there is no market substitution. Any harm for the outputs of AI, the DOJ argues, has to be assessed separately, under the Supreme Court’s recently enunciated Warhol framework. (“The same copying may be fair when used for one purpose but not another,” Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 533 (2023).)
Will This Filing Change Litigation Outcomes?
Statements of interest under § 517 may carry persuasive weight because they come from the government, but they are not binding, and judges are free to give them as much or as little deference as they choose. Techdirt and Above the Law both ran pieces recently arguing DOJ’s position is not disinterested because the Trump administration is reportedly negotiating an equity stake in OpenAI. But regardless of the brief’s ultimate influence on the outcome, the filing is a well-organized statement of the fair use argument for AI training, and therefore may be used as a playbook for other defendants, or for courts, to analyze the issue.
On the other hand, the political polarization of our country may mean that some judges will view the position of the brief negatively, merely because of its source–and the Trump administration’s track record of pushing the envelope on intervening in court process. But it is not clear this is a red versus blue issue. While the brief gravitates toward strong language about national security interests, its substance on fair use is probably what most intellectual property lawyers expect in the analysis of this issue.
