Governments do not usually pick sides in private copyright fights. This week the United States did. In a 20-page statement of interest filed Tuesday, the Justice Department told the federal court hearing the consolidated case against OpenAI that training large language models on copyrighted text is protected by fair use, siding squarely with the ChatGPT maker against the New York Times, a group of other newspapers, and a long roster of authors. It is the clearest signal yet of how far the current administration will go to keep American AI unencumbered.
The department's language was expansive. Training models on writing found across the internet delivers "creative possibilities and public benefits" that "far outweigh any competitive harm," it argued, and a ruling for the Times would "thwart creative and scientific progress while hindering American prosperity and economic mobility." Then it reached for the argument that now attaches itself to almost every AI dispute: national security. Rules that make it harder to build a strong AI industry, the filing warned, "threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered." Fair use, in this telling, is not just a copyright doctrine but a matter of keeping ahead of China.
The Times was unimpressed. The administration "is siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work they stole," spokesman Graham James said. His counter is narrower and harder to wave away: "AI companies simply need to pay fairly for the content that makes their products possible, as copyright law requires." The paper's original 2023 suit did not argue that AI should not exist, only that OpenAI was free-riding on a massive investment in journalism to build a product that competes with it. Ranged alongside the Times are the New York Daily News, the Chicago Tribune, Ziff Davis, the Center for Investigative Reporting, and authors from George R.R. Martin and John Grisham to Sarah Silverman and the Authors Guild.
What makes the intervention striking is less the position than the posture. The Justice Department was careful to note that its reasoning "apply similarly to all parties," including the book authors and publishers in the related cases, which means the government is effectively arguing a whole category of creators out of a remedy in litigation to which it is not a party. That is a heavy thumb on a civil scale, and it lands at a delicate moment for the defendant. OpenAI, despite billions spent and no profit to show, filed IPO paperwork with the SEC in June and is expected to go public soon. A federal endorsement of its core legal risk is worth a great deal to a company about to ask investors to price exactly that risk.
The case itself may pause before it resolves anything. On 31 August, Judge Sidney H. Stein ordered the Times to explain by 11 September why the action should not be stayed while summary judgment plays out in the other cases consolidated with it, with OpenAI's response due a week later. Whatever the court decides, the deeper question the filing raises will outlast it. If training on copyrighted work is fair use because the national interest requires it, then the people whose work feeds these systems are being asked to subsidise them, and told that the subsidy is patriotic.