DOJ Backs OpenAI in NYT Copyright Case: Fair Use

The federal government has, for the first time, taken a formal side in the wave of copyright lawsuits between publishers and AI companies — and it sided with the AI companies. On September 1, 2026, the U.S. Department of Justice filed a statement of interest with U.S. District Judge Sidney Stein in the Southern District of New York, urging the court to hold that training large language models on copyrighted written works qualifies as fair use under U.S. copyright law.

The filing lands in the consolidated litigation stemming from The New York Times’ 2023 suit against OpenAI and Microsoft, which alleges that millions of Times articles were copied without authorization to train systems including ChatGPT. The DOJ’s roughly 20-page brief does not make the government a party to the case, but it asks the court to reject the newspapers’ core theory of infringement outright.

What the DOJ argued

The government’s position is stated bluntly in the brief: “The United States has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law.” The DOJ frames the ingestion of text to train a model as a transformative use — one that learns statistical patterns from works rather than reproducing them — and argues that treating training itself as infringement would “distort” copyright doctrine.

The brief pairs that legal argument with an economic and geopolitical one. Requiring AI developers to license every copyrighted work in a training corpus, the DOJ argues, would erect barriers to competition, concentrate the technology in the hands of the few companies that can afford blanket licensing, and slow domestic AI progress. The government asserts the United States has a “profound national interest” in maintaining a dominant AI industry, warning that overly restrictive copyright rules could hand an advantage to foreign adversaries — a reference to Chinese labs that, the brief suggests, operate under looser intellectual-property constraints.

The upshot of the DOJ’s framing is a division of labor between the branches of government: courts should read fair use broadly, and any new compensation regime for creators should come from Congress, not from an expansive interpretation of the existing Copyright Act. It is a message that effectively tells rightsholders their remedy is legislative, not judicial.

What a statement of interest is — and isn’t

A statement of interest is a mechanism federal agencies use to put the government’s view on the record in a case they are not party to. It carries no binding authority. Judge Stein will decide independently whether OpenAI’s training and output meet the four-factor fair use test, and he is free to disregard the DOJ’s reasoning entirely.

Still, the filing is significant precisely because it is unusual. Legal observers described it as a rare intervention: the executive branch weighing in on a private commercial copyright dispute to advocate for one side’s reading of the law. Because the same fair use question sits at the center of parallel suits brought by authors, music labels, and other news organizations, a persuasive government brief in the marquee Times case could ripple across the entire docket — including actions like the one brought by The Intercept and other outlets that were folded into the same proceedings.

The New York Times pushes back

The Times reacted sharply. Spokesman Graham James 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,” and warned that letting companies “take that content without permission or compensation would undermine the sustainability of the human-created content that a healthy society depends on.”

That argument goes to the fourth fair use factor — the effect on the market for the original work. Publishers contend that models trained on their journalism and capable of summarizing or reproducing it compete directly with the outlets that produced it, eroding subscriptions and licensing revenue. It is the same market-harm theory underpinning the Sony and Warner suit against Anthropic over song lyrics and the copyright rulings beginning to emerge in other jurisdictions, such as the German court decision against Suno over training data.

The industry’s counter-strategy has increasingly been to license rather than litigate where it can. OpenAI has struck content deals of its own, including the Getty Images licensing arrangement, and rivals have signed similar pacts — an implicit acknowledgment that at least some training data is worth paying for, even as the companies argue in court that they are not legally required to.

The conflict-of-interest question

The filing arrives against an awkward backdrop. In July, OpenAI CEO Sam Altman floated the possibility of the federal government taking roughly a 5% stake in the company, a proposal detailed in reporting on a potential U.S. government equity position in OpenAI. Critics were quick to note the optics of the Justice Department advocating for OpenAI’s legal position while the administration is separately exploring a financial interest in the same company.

The DOJ’s brief does not mention any such discussions, and a statement of interest is not an endorsement of a party’s business — it is an argument about the law. But the appearance issue is unlikely to fade, particularly as AI companies pour record sums into Washington influence, a trend captured in the surge of OpenAI and Anthropic lobbying spending. For creators already convinced the deck is stacked, a government that both argues for and may invest in the industry is a difficult picture to counter.

Where the case stands

Procedurally, the litigation is still working toward a decision on the merits of the fair use defense. The DOJ’s filing does not resolve anything; it adds a heavyweight voice to OpenAI and Microsoft’s side of a question the court has yet to answer. Discovery in the consolidated cases has been contentious, including disputes over the preservation of ChatGPT logs and the extent to which the newspapers can probe how their content was used.

Judge Stein now weighs a fair use analysis with the executive branch openly urging one outcome — an input that is legally non-binding but politically loud.

What it means

The stakes here run well beyond one newspaper. If courts accept the broad fair use theory the DOJ is pressing, the foundational legal risk hanging over the entire generative-AI industry largely dissolves: labs could train on the open web and published works without licensing, and the leverage shifts decisively toward the companies. If courts reject it, the economics of building frontier models change overnight, forcing either expensive licensing regimes or narrower training corpora.

Who wins if the DOJ’s view prevails: the largest AI developers, who avoid a licensing bill that could run into the billions, and who gain a durable legal shield for models already trained. Who loses: publishers, authors, and rightsholders, whose primary source of leverage — the threat of infringement liability — would be sharply curtailed, pushing them toward voluntary deals negotiated from a weaker position or toward a Congress that has shown little appetite to act.

The national-security framing is the part to watch. By casting permissive copyright rules as a matter of competing with China, the government reframes a commercial dispute as an instrument of industrial policy — the same logic that has driven export controls and federal AI initiatives across the sector. That framing is powerful in a courtroom and even more so in Congress, and it suggests the administration will keep treating frontier AI as strategic infrastructure to be protected, not an industry to be constrained. The immediate question is narrow — is training fair use? The larger one is whether copyright, as a check on how AI systems are built, survives contact with national-security politics at all.