DOJ Files Statement of Interest in OpenAI Copyright Litigation

The Justice Department argues in court that training a large language model on copyrighted text is fair use, but outputs are a separate question.

DOJ filed the 20-page statement of interest on September 1 in In re OpenAI, Inc., Copyright Infringement Litigation (S.D.N.Y., MDL No. 3143, Stein, J.), under 28 U.S.C. § 517, which, unlike an amicus brief, requires no motion or leave of the court. The statement was filed for "All Matters," so the reasoning applies to the co-pending book-author and publisher cases alongside the NY Times case.

DOJ frames compulsory licensing as a competition problem: entry barriers would hand "the largest technology companies" an "oligopoly on LLM training" and subsidize "old mainstream media companies." A Times spokesperson said the administration is "siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole."

Key Arguments in the Filing

  • One use at a time. Citing Warhol v. Goldsmith, the government treats collecting, training, and generating as three separate uses, and argues fair use only as to training.
  • Use of copied material. The first fair-use factor asks whether the new use adds something different. Training converts prose into statistical patterns, which Judge Alsup called "transformative—spectacularly so" in the Anthropic case.
  • Lost sales. The fourth factor considers market harm, and whether the copying produces "significant substitutive competition." Training publishes nothing, so nothing substitutes.
  • Attack on Kadrey. In the Meta case, Judge Chhabria raised market dilution: LLMs could flood the market with competing books. DOJ calls that "contrary dicta" and "deeply flawed" for treating training and outputs as one use.

Potential Impacts

  • DOJ's statement is persuasive authority only. Judge Stein is not bound. Summary judgment motions are due today, September 4; replies November 6.
  • Market dilution is the strongest market-harm theory copyright owners have. If it holds, training licenses may become unavoidable. If DOJ's framing prevails, AI company exposure narrows to outputs that reproduce original works.
  • The same administration is widening patent eligibility for AI inventions under Director Squires while arguing copyright should not reach AI training. AI developers gain on both fronts: more of what they build is patentable, less of what they train on is protected.

The Third Circuit heard argument in Thomson Reuters v. ROSS Intelligence on June 11. DOJ filed nothing there. The first binding decision on AI fair use is thus poised to come from an appeal the government has not weighed in on.

Sources

  1. Reporting: Blake Brittain, Reuters (Sept. 2, 2026), https://www.reuters.com/legal/litigation/us-government-backs-openai-new-york-times-copyright-case-2026-09-02/
  2. Primary document: Statement of Interest of the United States, In re OpenAI, Inc., Copyright Infringement Litig., No. 1:25-md-03143 (S.D.N.Y. Sept. 1, 2026), Dkt. 1682, https://www.courtlistener.com/docket/69879510/1682/in-re-openai-inc-copyright-infringement-litigation/