The US Department of Justice just planted its flag in the most consequential copyright battle of the AI era. In a brief filed in Manhattan federal court, the DOJ argued that training large language models on copyrighted material qualifies as fair use, and that restricting it would threaten American prosperity, economic mobility, and national security.
The filing landed in the consolidated lawsuit between The New York Times and OpenAI/Microsoft, a case that has been grinding through the courts since December 2023. It marks the first time the federal government has formally intervened in any of the sprawling copyright disputes surrounding AI training data.
The government’s argument
The DOJ’s core claim is straightforward: when an LLM ingests copyrighted text during training, it is not copying that text in any meaningful competitive sense. The outputs do not reproduce or substitute for the original works. In the government’s framing, this process is “extraordinarily transformative,” a legal term of art that carries real weight in fair use analysis.
The brief went further than just defending the mechanics of AI training. It argued that clamping down on data access would “significantly hamper the progress of science and useful arts,” language that deliberately echoes the constitutional purpose of copyright law itself.
