DOJ files brief arguing AI training on copyrighted text is fair use in NYT v. OpenAI/Microsoft case

The US Department of Justice filed a brief in the consolidated NYT v. OpenAI/Microsoft copyright case, arguing that training large language models on copyrighted text is fair use because the copying done during training is legally distinct from what a model outputs. The brief also criticises the US Copyright Office's 2025 report, which had rejected the idea of a blanket fair use exemption for AI training. No primary source document (the actual court filing) was linked in the reporting reviewed; this account relies on secondary coverage from The Decoder.

This is a real filing in a bellwether case that many other AI copyright disputes are watching, so it signals where the executive branch currently stands on a question with billions of dollars in licensing and liability exposure for AI labs, publishers and content creators. A DOJ statement of interest is persuasive, not binding, so it does not resolve the case or change any company's legal obligations today, but it nudges the odds and gives OpenAI and Microsoft a notable ally in argument. Enterprises and developers relying on LLMs trained on scraped content should treat this as one data point in an unresolved legal environment, not a settled outcome.

Framing this as "landmark" or decisive overstates what a statement of interest actually does: it is the DOJ's opinion offered to the court, not a ruling, settlement or change in law. Coverage that implies this resolves the fair use question for AI training, or that it obligates courts to follow the DOJ's reasoning, is misleading. The absence of a link to the actual filing in the reporting reviewed is also a gap worth flagging, since the specifics of DOJ's legal argument matter more than the topline summary.

1. 01Whether the presiding judge cites or relies on the DOJ's statement of interest in any ruling on the fair use question in NYT v. OpenAI/Microsoft.
2. 02Whether DOJ or other federal agencies file similar statements in other pending AI copyright suits (e.g. against Anthropic, Meta, Stability), which would confirm a coordinated policy stance rather than a one-off filing.
3. 03Any response or revision from the US Copyright Office to its 2025 report given direct criticism from DOJ, and whether Congress (via Joe Morelle or others) reacts formally.