OpenAI gains US backing

Washington enters the AI copyright fight

OpenAI has gained a powerful new ally in its copyright fight with publishers: the US government. In a statement of interest filed on September 1 in the multidistrict OpenAI copyright litigation in the Southern District of New York, the Justice Department urged the court to reject the argument that training large language models on copyrighted text is, by itself, unlawful [1]. The filing supports OpenAI’s core position in cases brought by The New York Times and other publishers: that copying written works for model training can fall within fair use when the purpose is to build a system that learns patterns rather than to republish the original expression [1].

The intervention does not decide the lawsuit. A statement of interest is advisory, and the judge remains free to accept or reject the government’s reasoning [3]. But the move materially changes the legal and political setting of the case. What began as a dispute over news articles, licensing markets and alleged substitution is now being framed by Washington as a question of industrial policy and national security [1].

The Justice Department’s brief says the United States has a strong interest in sustaining a robust and competitive AI industry and in keeping American companies at the front of global AI development [1]. It argues that rules making LLM development significantly harder in the US could give foreign adversaries an advantage and could impair the deployment of AI for national-security uses, including intelligence analysis and other government functions [1].

The fair-use argument gets federal reinforcement

At the center of the dispute is whether training a model on copyrighted works is “transformative” enough to qualify as fair use. The Justice Department’s answer is emphatic: it says LLM training is “extraordinarily” transformative because the copying is used to teach statistical relationships in language rather than to sell or display the original articles [1]. The filing also stresses a use-by-use distinction: training a model and generating an output that reproduces protected expression are separate legal questions [1].

That distinction is crucial for OpenAI. Publishers argue that AI companies used journalism without permission or payment to build commercial products that can compete with the original news providers [2]. The Times first sued OpenAI and Microsoft in 2023, accusing them of using millions of its articles without authorization to train ChatGPT and related systems [3]. Other news organizations, including MediaNews Group-owned papers, Ziff Davis and the Center for Investigative Reporting, have joined related litigation [2].

The government’s brief does not say every AI output is lawful. It recognizes that outputs reproducing protected expression may raise separate questions, while insisting that those questions should not determine whether the training process itself is fair use [1]. This gives OpenAI a clearer doctrinal line to argue: if a model occasionally produces infringing text, the remedy should address that output, not outlaw the underlying training practice.

Why the government says publishers could hurt smaller players

One of the brief’s most consequential claims is economic. The Justice Department argues that a ruling requiring broad licenses for training could entrench the largest AI developers because only the best-capitalized firms would be able to afford the necessary content deals [1]. It also says licensing fees would disproportionately benefit legacy media companies with large archives, while potentially raising barriers for start-up AI firms and smaller publishers [1].

Nieman Lab reported that the filing casts a publisher win as a threat not only to AI companies but also to independent and start-up media, because the government argues that LLMs can help smaller newsrooms compete with better-resourced outlets [4]. The brief points to uses such as assisting authors with production tasks or directing users toward a wider set of sources, including dissenting or less dominant perspectives [1]. That argument is controversial because it turns a claim about uncompensated use of journalism into an argument that unfettered AI training might benefit journalism’s smaller players.

The policy implication is large. If the court accepts the publishers’ position, data access could become a major cost center for frontier AI development, and the market for training licenses could become a new revenue stream for media companies. If OpenAI’s position prevails, current data-intensive model-building practices would gain legal support, and publishers would likely have to focus their claims on infringing outputs, contractual restrictions or specific forms of substitution rather than training alone.

Publishers push back

The Times strongly rejected the administration’s stance. A spokesperson for the paper said the government was siding with a small number of extremely valuable AI companies at the expense of creators whose work had allegedly been taken [2]. The Times also argued that AI companies can thrive while paying fairly for the content that makes their products possible [2].

The response goes to the heart of the media industry’s concern. Newsrooms invest heavily in reporting, editing, legal review, photography, audience development and archives. Their fear is that AI systems can absorb that investment, reduce referral traffic and offer users synthesized answers that weaken the business case for visiting or subscribing to the original publisher [2]. AP reported that The Times has argued OpenAI threatens journalists’ livelihoods by using billions of dollars’ worth of its work and, in some cases, allegedly producing Times material verbatim in response to user prompts [2].

Wired also reported criticism from other rights-holder groups, including the Authors Guild, whose chief executive described the government’s letter as deeply disappointing and legally flawed [5]. Those objections reflect a broader coalition of publishers, authors, music companies and other copyright owners that has filed cases against AI developers over training data [3]. For those plaintiffs, the danger is not merely one model or one defendant; it is the possibility that courts normalize mass ingestion without permission as the default foundation for generative AI.

Why investors and developers are watching

For AI investors, the filing helps quantify a key uncertainty: legal exposure around training data. A broad loss for OpenAI could raise expected costs through licensing, damages, settlements or constraints on model development. A win could reduce that risk and validate the premise that large-scale ingestion of text for training is a legally protected input to innovation.

Bloomberg Law reported that the Justice Department urged the judge to rule for OpenAI’s fair-use theory and said the case could guide future rulings over what counts as infringement in AI model training [6]. Reuters similarly reported that this appears to be the first time the US government has weighed in on the wave of AI copyright cases brought by authors, publishers, music labels and news outlets [3]. That makes the brief important even if it is not binding: it supplies a federal policy template for other courts and for future lobbying over AI and copyright.

The timing also matters. Courts have begun to split on how to analyze AI training, especially on market harm and whether training should be separated from output behavior [3]. The Justice Department sides with the view that the training step is a distinct, highly transformative use and that generalized competition from AI-generated material should not automatically count as cognizable copyright harm [1].

What happens next

The immediate question is whether Judge Sidney H. Stein gives the filing meaningful weight in the OpenAI multidistrict litigation. The government has no power to end the case by filing its view, but judges often take federal statements of interest seriously, especially when they address national policy, statutory interpretation and market-wide consequences [5].

The broader question is whether copyright law will become a licensing engine for AI training or a narrower tool aimed at outputs that reproduce protected expression. The US government has now placed itself firmly on the second side of that divide. For OpenAI, that is a significant strategic gain. For publishers, it is a warning that their battle is no longer only against a technology company, but against a federal policy preference for rapid AI development.

The court still has to decide the law. But Washington has made clear where it thinks the balance should fall: protect creators from infringing outputs, encourage licensing where parties choose it, and avoid a rule that would make permission from every major content owner a precondition for building frontier AI systems [1].

Sources from the last 72 hours

1. [1]Statement of Interest of the United StatesSep 1, 2026, 12:00 AM UTC
2. [2]US government backs OpenAI in New York Times copyright caseSep 2, 2026, 10:53 AM UTC
3. [3]Trump administration says a New York Times win in OpenAI lawsuit would “threaten national security” and hurt small newsroomsSep 2, 2026, 8:22 PM UTC
4. [4]Trump Administration Sides With OpenAI in New York Times Copyright LawsuitSep 2, 2026, 2:41 PM UTC
5. [5]Trump Administration Backs OpenAI in NY Times Copyright SuitSep 2, 2026, 7:44 PM UTC
6. [6]Trump administration backs OpenAI in New York Times’ copyright case over training of chatbotsSep 2, 2026, 7:12 PM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.