Trump's Justice Department tells a federal court that training AI on copyrighted news is fair use, backing OpenAI against The New York Times

The U.S. Department of Justice told a federal judge on September 2 that training AI systems on copyrighted news articles is legal under fair use -- the first time the federal government has formally intervened in the wave of copyright lawsuits against AI companies, and a direct assist to OpenAI in its highest-profile fight. Associate Attorney General Stanley Woodward Jr., who signed the filing, called it "a historic statement of interest" in a post on X, adding that President Trump "has made clear that AI dominance is critical to promote national security, prosperity, and economic mobility for all Americans."

The filing lands in the consolidated docket In re: OpenAI, Inc., Copyright Infringement Litigation (MDL No. 3143), the Southern District of New York case that folds The New York Times' 2023 lawsuit together with more than a dozen others against OpenAI and its partner Microsoft -- including suits from the Daily News, the Center for Investigative Reporting, and the Ziff Davis network of sites -- all assigned to U.S. District Judge Sidney H. Stein. Most coverage of the filing led with the Times alone; the government's brief actually speaks to the fair-use question for the whole consolidated docket at once.

What the government actually argued

The DOJ's core claim is that using copyrighted text to train a model is transformative -- the model doesn't reproduce the work, the argument goes, it learns statistical patterns from it -- and that transformative use sits squarely inside fair use as courts have long defined it. "The United States has a strong interest in this court rejecting any argument that training LLMs on copyrighted texts violates copyright law," the filing states, framing the stakes as competitive rather than purely legal: forcing AI companies to license training data at scale, the government argued, would be "legally incorrect" and could hand the advantage to rivals operating under looser rules.

Woodward put the same argument more bluntly outside the filing itself: "This Administration will never let our Nation be at a disadvantage relative to our foreign adversaries based on a plainly incorrect understanding of copyright law."

The Times rejected the government's framing outright. "The Administration is siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole," said Times spokesperson Graham James, arguing the government's position would "undermine the sustainability of the human-created content that a healthy society depends on, and which AI needs to function."

The pushback from inside the same docket

The sharpest critique came from a lawyer representing a different plaintiff in the same consolidated case. Matt Topic, an attorney for The Intercept -- one of the more than a dozen outlets whose suits were folded into MDL No. 3143 -- argued the DOJ's brief "isn't based on any understanding of actual facts," since the department was never a party to the two years of discovery the publishers have conducted. He called the administration's position "out of touch and a threat to the financial survival of media outlets."

The brief doesn't land in a vacuum. Anthropic agreed in May to pay $1.5 billion to settle a related case after a judge found it had illegally downloaded pirated books to train its models -- the closest thing the industry has to a fair-use price tag, and a settlement rather than a ruling precisely because Anthropic chose not to test in court the question the DOJ is now trying to answer for OpenAI. A coalition of roughly 400 news publishers sued OpenAI and Microsoft on similar grounds in June, and musicians sued the AI-music startup Suno this week over the same underlying theory applied to audio. Every one of those cases turns on the question the DOJ's filing tries to settle in one motion: does training itself infringe, regardless of what a chatbot's output looks like?

What's next

Judge Stein has given both sides in the consolidated docket until Friday, September 4 to file for summary judgment -- the point at which a judge can rule on the core legal question without a trial, if the underlying facts aren't genuinely in dispute. A DOJ statement of interest carries no independent legal weight; Stein can credit it, ignore it, or weigh it alongside everything else, the way judges typically treat amicus-style filings. But it now sits in the record the same week both sides ask him to decide the case, putting the executive branch on record making the argument OpenAI has been making alone for nearly three years.

"An unprecedented, uncompensated transfer of IP rights from news organizations to tech companies." -- Matt Topic, attorney for The Intercept in the consolidated copyright docket

For a defendant three years into the same lawsuit, having the Justice Department volunteer its argument for free is close to the best outcome short of a plaintiff walking away. But a statement of interest is not a ruling, and Judge Stein's own March 2025 order already found the Times' core copying claims strong enough to survive dismissal once. Whether "transformative" training beats that finding a second time is now his call, on a record the government just added itself to.