Statement of Interest of the United States on fair use in In re OpenAI, Inc. Copyright Infringement Litigation (S.D.N.Y., 1 September 2026)

Key points

On 1 September 2026 the United States told the court that copying written works to train large language models is, in general, fair use.
The filing confines itself to the training stage, which it calls exceedingly transformative, and takes no position on acquisition of the works or on particular outputs.
It is a statement of interest under 28 U.S.C. 517, not a ruling: it binds no court but records the executive branch's position, which later briefs will cite.

Summary

On 1 September 2026 the United States filed a 20-page Statement of Interest in the consolidated copyright cases against OpenAI in the Southern District of New York (25-md-3143, Document 316 in the lead member case, signed by the Associate Attorney General and the head of the Civil Division). Appearing under 28 U.S.C. 517, the government says it "has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law", argues that "the use of copies to train LLMs is extraordinarily transformative", and frames the stakes as competitiveness and national security: rules that hamper US AI development "threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered". It is explicit about what it does not decide. The government "focuses on the question whether the use of copyrighted works at the training stage" is fair use, concedes that "at the output (rather than training) stage, certain uses may not be transformative if the LLM reconstructs and disseminates an original copyrighted work", and takes no position on how the works were obtained. For animation education the document matters as the first formal executive-branch position on the question that Andersen v. Stability AI and the studios' case against Midjourney turn on, even though it concerns written works and language models rather than images or video.

Implications for animation education

Nothing in the filing changes what a program or a student may lawfully do today; it is advocacy addressed to one court in one set of cases about text. Its practical effect for teaching is on the argument map. In an IP or ethics unit the fair-use question now has a formal executive-branch position on the training side, with a stated reservation on outputs, and both halves are quotable: the government's case for training rests on the copies never being made "accessible to the public" and on outputs "lacking in substantial similarity" causing no cognizable market harm, which is exactly the ground on which image and video cases are fought. Course leads should present it as one party's brief, note that it disagrees with a Northern District of California decision (Kadrey v. Meta) and with the Register of Copyrights' Part 3 report, and watch for how the Andersen and Midjourney courts treat it. Students working from image models should not read "training is fair use" as settled law for their medium.

Source

Source: United States Department of Justice ↗ (Policy Text)

Primary document: storage.courtlistener.com/recap/gov.uscourts.nysd.640396/gov.uscourts.nysd.640396.1682.0.pdf ↗

Cite this item

United States Department of Justice (2026). ‘Statement of Interest of the United States on fair use in In re OpenAI, Inc. Copyright Infringement Litigation (S.D.N.Y., 1 September 2026)’, United States Department of Justice. Available at: https://storage.courtlistener.com/recap/gov.uscourts.nysd.640396/gov.uscourts.nysd.640396.1682.0.pdf

Download .ris Download .bib

Your reference manager can also read this page directly: with the Zotero (or Mendeley) browser connector installed, save it straight to your library. Whole-collection exports: RIS, BibTeX, CSL-JSON.