The United States filed a 20-page statement of interest in The New York Times v. OpenAI, arguing that constraining large-language-model training under a "misunderstanding of fair use doctrine" would thwart progress. [1] "The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally," the brief says, citing President Donald Trump's January 2025 executive order on retaining global AI leadership. [1] This is a Wednesday filing. It is not a ruling.
Private plaintiffs pressed the other side of the copyright vise last week, when Sony and Warner Chappell sued Anthropic as a different defendant. Today's receipt puts the executive branch on OpenAI's side against a named publisher. Same-day G20 remarks in Chapel Hill have the administration telling the world it will not take a heavier hand on AI releases. The brief is a heavier hand for the lab. The Tumbler Ridge safety suits are a separate file.
TechCrunch treated the intervention as a fair-use and legal-process story. [1] The authors of the brief lack jurisdiction in the Southern District of New York. The case is being tried there. "This intervention by the Trump administration could still carry weight," Amanda Silberling wrote. [1] A statement of interest is not a judgment. A reader who only saw "government sides with OpenAI" missed that the court has not decided anything.
The brief's purpose clause is global leadership. The named opposing interest is the Times. The fair-use fight, TechCrunch writes, turns on whether training on copyrighted books, articles and other media without permission is "transformative" enough to be legal. [1] "Constraining LLM development under a misunderstanding of fair use doctrine would thwart such creative and scientific progress while hindering American prosperity and economic mobility," the brief says. [1] That is the administration's sentence. It is not the court's.
Last year's Anthropic settlement, TechCrunch notes, fined the company $1.5 billion for pirated shadow-library copies, not for the training itself. Judge William Alsup compared the models to a reader aspiring to be a writer: they trained "not to race ahead and replicate or supplant" the works "but to turn a hard corner and create something different." [1] That is a different docket and a different defendant. OpenAI is the lab named in SDNY. The Times is the publisher. Wednesday's record is the executive choosing a side. ChatGPT, Claude and Gemini all train on "incomprehensibly massive databases," Silberling writes. [1] The brief still names one lawsuit. The publisher is the named cost.
The Times has already accused OpenAI of hiding evidence in the ChatGPT copyright trial, TechCrunch notes in a linked July file. [1] That allegation is not today's filing. Today's filing is the United States walking into the publisher's case to argue that a "misunderstanding of fair use" would block American AI leadership. Silberling says cases so far have largely favored the labs. [1] A pattern in other courts is not a ruling in this one. The authors of the 20-page brief do not sit in SDNY. They can still be heard there. That is the whole of Wednesday's receipt.
-- SAMUEL CRANE, Washington