The US Department of Justice has filed a "statement of interest" with a federal court in Manhattan, siding with the defendants in The New York Times' lawsuit against OpenAI and Microsoft. The filing's central claim fits in one sentence: training large language models on copyrighted works constitutes fair use.
The case is before Judge Sidney H. Stein of the US District Court for the Southern District of New York. The Times sued in late 2023, alleging the two companies used its journalism without permission to train ChatGPT, and both sides have since filed motions for summary judgment, with Stein setting this Friday as the filing deadline. The DOJ's timing was no accident.
The argument hinges on one word: transformative
The DOJ's reasoning runs like this: the training process extracts generalized relationships from source material, and the model itself does not distribute that source material, making the use highly transformative. The filing goes further, tying AI development to scientific progress, national security and US economic competitiveness, arguing the resulting benefits far outweigh any competitive harm to the market for the original works.
Principal Deputy Associate Attorney General Stanley Woodward Jr. called the filing "a historic statement of interest," invoking the president's own assessment that "AI dominance is critical to promote national security, prosperity, and economic mobility." The same day, US officials at the G20 called on other countries to leave room for AI training while still protecting creators — a consistent line from Washington.
What it can do, and what it can't
A statement of interest is not a ruling, and it does not bind the court. What it offers is the executive branch's policy view, which can shape how the judge frames the analysis; it is no substitute for weighing the four fair-use factors one by one — especially the fourth, the effect of the use on the potential market for the original work, which is precisely where news publishers have their strongest, and most plausible, path to a reversal. The AI industry as a whole does not get a copyright exemption just because of one brief, either.
The Times' response, delivered by spokesperson Graham James, was pointed: it accused the administration of siding with "trillion-dollar AI companies" in a way that would "undermine the sustainability of the human-created content that a healthy society depends on."
The spillover will arrive before the verdict does
There's already a reference point: Anthropic settled a book-copyright case this year for $1.5 billion, after a judge found it had illegally downloaded copyrighted books. That case turned on how the data was obtained — pirated book sources — a question separable from the act of training itself. The DOJ's filing likewise sidesteps data provenance, addressing only the nature of the training step.
That framing will decide how bargaining power gets split at the negotiating table over the next few years. If courts adopt a "training equals transformative use" framework, what the music, publishing and news industries can claim narrows to a single question — whether the method of acquiring the data was lawful — shrinking the scope of potential damages and loosening the pricing basis for licensing fees. One layer of spillover is already visible: domestic buyers of Chinese-language text corpora and film and music rights have long benchmarked their offers against overseas licensing deal prices; once training defaults to fair use in the US, sellers' pricing anchor is likely to drift downward.
No ruling has landed yet. The next milestone is the pair of summary judgment motions now sitting with Judge Stein.
Sources: US Department of Justice statement of interest, Southern District of New York public court docket, CocoLoop, GV Wire, DataStudios; the judge's name, the parties to the case and the $1.5 billion settlement figure were cross-checked against two independent public reports.