The Justice Department told a Manhattan federal court that training large language models on copyrighted texts should not be treated as a copyright violation, according to a 20-page statement of interest filed late Tuesday, Sept. 1, 2026, and reported Wednesday. It is not a ruling. The New York Times' lawsuit against OpenAI and Microsoft is still pending in the Southern District of New York.

The United States appeared under 28 U.S.C. § 517 in In re OpenAI, Inc. Copyright Infringement Litigation, No. 25-md-3143, a multidistrict proceeding before Judge Sidney Stein that includes the Times' December 2023 complaint, which seeks billions of dollars in damages from OpenAI and Microsoft, The Verge reported. The government said it 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," and quoted President Donald Trump's Jan. 23, 2025, executive order on "Removing Barriers to American Leadership in Artificial Intelligence," including the line that the United States must "retain global leadership in artificial intelligence."

"The New York Times seeks to narrow fair-use doctrine to exclude the training of OpenAI's large language models (LLMs)," the brief states. "That result would be inconsistent with basic copyright law principles and severely hamper 'the Progress of Science and useful Arts.'" It adds: "Constraining LLM development under a misunderstanding of fair use doctrine would thwart such creative and scientific progress while hindering American prosperity and economic mobility." The President's National Policy Framework for Artificial Intelligence, cited in the filing, says "the training of AI models on copyrighted material, in and of itself, does not violate copyright laws."

The United States "focuses on the question whether the use of copyrighted works at the training stage — by copying works in order to feed data into the model as learning material — constitutes fair use." Training, it argues, is "exceedingly transformative," does not show the original work to the public, and does not substitute for Times articles in the way copyright measures market harm. Output that reconstructs protected expression, the brief says, is a different use. "Rules of law that make it significantly more difficult to develop a robust AI industry in the United States therefore threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered," the filing states. "It would be problematic — and legally incorrect — to impose broad copyright liability that would generally render training of AI models impermissible without licensing."

A Times spokesperson told The Hollywood Reporter the administration is "siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole." "Both AI and creators can thrive – AI companies simply need to pay fairly for the content that makes their products possible, as copyright law requires," the spokesperson said. "The Administration's proposal to let companies take that content without permission or compensation would undermine the sustainability of the human-created content that a healthy society depends on, and which AI needs to function."

Copyright is a statutory grant, not a veto over every new machine that reads. Fair use is the valve Congress left so that grant cannot lock up later tools. The Times wants a license gate on training. The Justice Department wants the opposite default, and it is using national-security and industrial-policy language to get it. Both are bids to have the government pick a side in a private fight. The brief itself treats a compulsory licensing regime, if the country wants one, as a job for Congress — not a rewrite of § 107 by this court.

Other courts have already split training from piracy. In 2025, Judge William Alsup found Anthropic could train on lawfully acquired books but still face liability for using shadow libraries — a track that produced a $1.5 billion settlement with authors, TechCrunch and The Verge reported. That is not this case, and Wednesday's filing does not close it.

Sources