DOJ files first US government brief backing AI fair use for training, contradicting White House framework
The Department of Justice put the full weight of the federal government behind AI companies in a major copyright dispute on September 1, 2026. The filing also exposed a split inside the Trump administration: the DOJ argued in federal court for exactly the outcome that the White House's own March 2026 AI policy framework had called a still-contested question.
What
The DOJ filed a 20-page statement of interest on September 1, 2026, in In re OpenAI, Inc. Copyright Infringement Litigation (MDL No. 25-md-3143) before US District Judge Sidney Stein in the Southern District of New York. The consolidated case covers the New York Times lawsuit and roughly 140 similar suits filed against OpenAI, Anthropic, Meta, Google, and others since 2023, per PYMNTS.
It is the first time the US government has intervened in any AI training copyright case, per Music Ally.
The central argument, per IPWatchdog's analysis of the filing, is that model training is "exceedingly transformative" because it converts text into numerical representations rather than reproducing works for their original purpose of informing or entertaining readers. The DOJ stated that "the United States has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law." The brief cited Google v. Oracle and Authors Guild v. Google as precedent.
The filing drew a three-stage line through the AI pipeline: acquiring training material, training the model, and generating outputs. The DOJ defended only the training stage. Claims about data acquisition practices and output-stage reproduction of protected expression remain live questions the brief left untouched, per Music Business Worldwide.
National security featured prominently. The DOJ warned that broad copyright liability would "create entry barriers benefiting only large technology companies," concentrate the LLM market, and weaken US economic competitiveness and national security capabilities, per IPWatchdog. The brief cited President Trump's executive orders from January 2025 and June 2026 on AI as the policy anchors for that position.
Why it matters
For teams building on AI APIs, the DOJ's intervention is the most consequential government signal yet on training-data liability. If Judge Stein adopts the DOJ's reasoning, the licensing-for-training argument that music labels, book publishers, and news organizations have been advancing would lose its most viable legal pathway.
The filing does not settle anything. A statement of interest carries no judicial authority on its own, and courts are not required to defer to executive branch legal positions. But the government's national-security framing raises the political stakes for plaintiffs. Arguing against fair use now means arguing against an official position that equates AI licensing requirements with strategic vulnerability to foreign adversaries, per the analysis at IPWatchdog.
The three-stage framework matters most for AI product builders. The DOJ did not defend data acquisition or output generation. Two of the three stages in a typical AI pipeline remain legally exposed. Companies whose web-scraping practices are challenged as unlicensed access, or whose outputs reproduce protected content substantially, get no cover from this filing.
For music-industry defendants, the picture is mixed. Anthropic faces suits from Universal Music Group, Concord, and ABKCO. Suno and Udio each face suits from Sony and Universal Music Group. Music Business Worldwide notes that the DOJ brief is technically limited to written works, but its training-stage reasoning applies directly to those cases. Music publishers had argued that output similarity proves training required licenses. The DOJ filing attacked exactly that reasoning, calling a prior ruling that linked training and output analysis "deeply flawed."
Context and reactions
The internal administration split is the story within the story. The White House's March 2026 National Policy Framework for Artificial Intelligence described the relationship between AI training and copyright as a still-contested question and directed agencies to study opt-out mechanisms for rights holders. The DOJ's September 1 filing asked a federal court to resolve that question in AI companies' favor rather than waiting for the interagency process the March document contemplated, per reporting covered in AOL News.
The US Copyright Office added a second layer of tension. Its May 2025 report declined to give a categorical answer on AI training fair use and called the question one for Congress to resolve. The DOJ brief acknowledged the Copyright Office's position only in a footnote, per Above the Law.
Above the Law also reported an undisclosed conflict of interest: the filing does not mention that the Trump administration has reportedly been in discussions about accepting an approximately $42.6 billion equity stake in OpenAI, one of the defendants whose case the brief advances. OpenAI had not confirmed or denied those discussions as of publication time.
The New York Times condemned the DOJ's intervention. Per the Times, as reported by PYMNTS, the administration was prioritizing "trillion-dollar AI companies" over creator compensation and the financial sustainability of journalism. Music Tech Policy, cited by Music Ally, described the approach as policy-driven and noted that courts do not automatically defer to White House positions on contested legal questions.
What to watch next
Three signals will determine whether the DOJ's position reshapes the litigation landscape. First, whether Judge Stein solicits responses from the Copyright Office or other agencies. Second, whether the White House issues any statement reconciling the DOJ brief with the March framework's more cautious language. Third, whether the DOJ files parallel statements of interest in the music-industry cases against Suno, Udio, and Anthropic, which would signal that the training-stage fair-use argument is now an official government-wide template rather than a single-case intervention.
Congress has not yet passed legislation on AI training and copyright. Until it does, the question will be resolved case by case in courts where the DOJ's brief will carry real weight even without binding authority.
Sources
- DOJ Sides with OpenAI, Warns Obstacles to AI Development Threaten National Security: IPWatchdog, analysis of the September 1 filing
- US DoJ says training AIs on copyrighted material should be fair use: Music Ally, September 3, 2026
- Trump administration sides with OpenAI: what it means for music's fight with Anthropic and Suno: Music Business Worldwide
- DOJ Tells Court AI Training Is Fair Use, Forgets To Mention It Is Negotiating A Stake In OpenAI: Above the Law, September 2026
- DOJ Sides With OpenAI Against the NY Times in High-Stakes Copyright Case: PYMNTS
- Trump Admin Splinters Over AI's Right To Train On Americans' Copyrighted Work: AOL News, September 7, 2026
