The Justice Department supports OpenAI in the copyright dispute with publishers.

      The Justice Department submitted a statement of interest supporting OpenAI’s fair use argument in the combined copyright cases against publishers, asserting that licensing fees could endanger national security. Unlike the U.S., EU law does not recognize a fair use doctrine, and the AI Act extends its copyright responsibilities to any provider offering a model in the Union market, regardless of where the training took place.

      The U.S. government has informed a New York court that utilizing publishers’ content to train AI models falls under fair use. The Justice Department filed this statement on Tuesday, according to CNET.

      The rationale given emphasizes national security concerns. Regulations that hinder the establishment of a strong American AI sector “threaten national security and provide an advantage to foreign competitors who do not face the same restrictions,” as stated in the filing.

      Moreover, the statement argues that licensing fees would effectively create an oligopoly on model training for the largest technology firms.

      The New York Times contests this view. Spokesperson Graham James remarked that the administration appears to favor a few trillion-dollar AI companies, asserting that these companies should fairly compensate for the content that enables their products.

      The lawsuits are consolidated in the Southern District of New York. The Times initiated its lawsuit in December 2023, followed by Ziff Davis, the parent company of CNET, in 2025, along with actions from 400 local newspapers.

      Such reasoning is not applicable in Europe, where EU copyright law features no fair use doctrine, only a limited set of exceptions.

      Instead, Europe has a text and data mining exception with an opt-out. Under Article 4(3) of the 2019 copyright directive, rights holders can maintain their rights, which then nullifies the exception.

      The AI Act directly connects to this framework. Article 53 mandates that every general-purpose model provider must implement a policy that identifies and honors these reservations.

      This obligation pertains to the model rather than the training process. Recital 106 clarifies that any provider launching a model in the Union market must comply with these rules, irrespective of where the copyright-relevant actions occurred.

      This recital outlines its own intent. No provider should be able to secure a competitive edge in the Union market by adhering to lower copyright standards than those in Europe.

      The Justice Department's argument can be interpreted in reverse. A Munich court ruled against OpenAI in November, determining that the memorized lyrics in GPT-4 constituted reproduction, and the mining exception did not apply.

      That ruling is currently under appeal and may be brought before the Court of Justice. It was made regarding actions that the Justice Department now claims are protected.

      Thus, a favorable outcome in Manhattan would not necessarily translate. OpenAI would still need to comply with the same European obligations in a market where publishers have previously accused it of withholding evidence.

       Published September 2, 2026 - 10:10 pm UTC

       Back to top