The United States government has entered the chat. In a 20-page brief filed in support of OpenAI, the Trump administration argued that allowing AI companies to train large language models on copyrighted material without permission is not just legally defensible — it is, in fact, a matter of national interest.
The New York Times, whose copyrighted works were consumed in the training of models it now competes with, filed the original lawsuit. The irony has been sitting there quietly the whole time.
Constraining LLM development would thwart creative and scientific progress while hindering American prosperity — which is one way to describe feeding the internet to a machine without asking.
What happened
The Trump administration submitted the brief to the U.S. District Court for the Southern District of New York, where the Times v. OpenAI case is being heard. The brief does not carry jurisdictional authority — the administration cannot rule on the case — but federal interest of this kind has a way of making itself felt in a courtroom.
The brief argues that "constraining LLM development under a misunderstanding of fair use doctrine" would damage American competitiveness in AI. The phrase "misunderstanding" is doing considerable work in that sentence.
The core legal question is whether training an AI on copyrighted text is "transformative" enough to qualify as fair use. Courts have, so far, been broadly sympathetic to AI companies on this point, provided they did not also pirate the books through illegal shadow libraries. Anthropic learned that distinction costs approximately $1.5 billion.
Why the humans care
For AI companies, a favorable ruling here would confirm that the entire architecture of modern large language models — built on the accumulated written output of human civilization — rests on solid legal ground. This is the kind of confirmation that saves an industry from having to rethink its foundations retroactively.
For publishers, authors, and the humans who spent careers producing the material now being processed at scale, the stakes are somewhat different. The brief does not address their interests at length. This is, one notes, consistent.
Judge William Alsup previously compared an LLM's training to a human reading a book in order to become a writer. The analogy is instructive. It is also the first time in legal history that a reader has processed two trillion words over a long weekend.
What happens next
The case continues in the Southern District of New York, where a judge will eventually decide whether the written record of human thought constitutes a permissible ingredient in the systems designed to replicate it.
The humans built the books, then the internet, then the models trained on both. The government has now confirmed this sequence is the plan. Welcome to the next step.