The DOJ filed a brief backing OpenAI's fair-use defense against The New York Times

The U.S. Department of Justice filed a 20-page Statement of Interest on September 1, 2026, in the Southern District of New York, in the long-running copyright case The New York Times brought against OpenAI and Microsoft. The filing doesn't make the DOJ a party to the case -- it's submitted under the statute that lets a federal agency lay out its legal position in someone else's litigation -- but it asks the court to rule that training large language models on copyrighted text is fair use.

The argument, in the government's own words

The brief frames the stakes as bigger than one publisher's claim: the DOJ says it has "a strong interest in continuing to develop a robust and competitive artificial intelligence industry," and warns that requiring licenses for training data would create a barrier only the largest technology companies could absorb -- effectively locking in an oligopoly rather than protecting competition. The Times pushed back through spokesperson Graham James, who said the administration's position would "undermine the sustainability of the human-created content that a healthy society depends on."

Not the same case as Anthropic's settlement

This is worth separating clearly from the Bartz v. Anthropic settlement covered here in August: that case resolved a claim that Anthropic built a training library out of pirated books, and the $1.5 billion payout was priced against that piracy specifically, not against training on lawfully acquired text. The Times' case against OpenAI and Microsoft is the other half of the same underlying legal question this site has been tracking -- whether training an LLM on lawfully obtained but copyrighted material is itself infringement -- and the DOJ's filing is the federal government's first stated position on that specific question, not a ruling, verdict, or settlement of any kind. The judge still has to decide.

For a company thinking about its own AI spend, the practical stakes are more concrete than the legal theory: if licensing becomes a mandatory cost of training a competitive model, that cost gets passed down through every vendor's pricing eventually. A government brief arguing against that outcome doesn't settle the question, but it's a real signal about which way federal policy is currently leaning on a cost that would otherwise show up, eventually, on every enterprise AI invoice.