The US government steps into the copyright litigation to defend fair use in AI training

On 1 September 2026 the United States Department of Justice filed a Statement of Interest with the District Court for the Southern District of New York, intervening without being a party in the consolidated proceeding In re: OpenAI, Inc. Copyright Infringement Litigation. In the twenty-page document, signed among others by Associate Attorney General Stanley E. Woodward Jr., the administration argues that reproducing protected works in the training phase of language models alone amounts to fair use under US copyright law. The filing draws an analytical line between training proper, the initial acquisition of the data, and the generation of output for the public, noting that each step raises distinct legal questions and that the federal intervention is strictly confined to the training process.

To support that claim, the government offers economic and national-security arguments. Citing an April 2022 Government Accountability Office report on the use of AI in intelligence analysis and weapons systems, the Department of Justice argues that rules capable of hobbling the US industry would benefit foreign adversaries operating without comparable constraints. On the market side, the filing frames any licensing mandate as a barrier to entry that would favour an oligopoly of well-capitalised players, working in practice as a subsidy to incumbent mainstream publishers. In footnote 13, however, the government narrows its own analysis, acknowledging that licensing deals for real-time, paywalled or proprietary content have already been signed by mainstream and independent publishers alike, regardless of fair use. The filing itself reports that the administration has encouraged Congress to consider licensing frameworks or collective management systems for rightsholders, while asking that any such law not settle when or whether a licence is required.

The response came through New York Times spokesperson Graham James, who said that allowing content to be used without permission or payment undermines the sustainability of the human work that artificial intelligence itself depends on. As industry coverage records, this is the first formal intervention by the US federal government in the copyright litigation over model training. The filing invokes two executive orders as its policy foundation: the 23 January 2025 order on removing barriers to American leadership in AI, and the 2 June 2026 order on innovation and the security of advanced AI. It is not known whether or when the court will rule on the point, nor whether the government will intervene in the other pending cases. The decision rests with the judge: a statement of interest does not bind the court and does not declare every training practice lawful. — Olya