The DOJ told a federal court that training LLMs on copyrighted text is fair use.

The Trump administration filed a brief September 1 in Manhattan federal court supporting OpenAI against the New York Times and other newspapers, arguing that "training of LLMs on written works is exceedingly transformative" and that the United States has a strong interest in the court rejecting any argument it violates copyright. TechCrunch reports this appears to be the first federal-government filing in the wave of AI training cases brought by authors, publishers, labels and news outlets. It doesn't bind the judge or decide the case, but it puts institutional weight behind the legal theory every model developer relies on.