The courtroom was quieter than a newsroom after deadline. I watched a terse DOJ filing land on a judge’s desk and felt the room tilt. You can sense that this case will change who pays for the facts we read and who builds the machines that read them.
I’ve covered copyright fights and technology for years, and you should know the basics fast: the Justice Department just sided with OpenAI in a copyright battle brought by The New York Times. That alignment—between a president who has publicly loathed the Times and an AI industry racing to scale—is both legal strategy and political theater.
The filing arrived Tuesday and the Justice Department framed a national-security argument first.
The brief argues that training large language models on written works is “extraordinarily transformative,” and that blocking such training could weaken U.S. competitiveness. I read the brief and the message is clear: the administration wants American labs—OpenAI, Microsoft and others—to keep innovating here, not overseas.
Associate Attorney General Stanley Woodward even posted on X that the administration won’t let the nation fall behind foreign rivals. Reuters and Gizmodo both flagged this as the federal government’s first sizable public stance in the many copyright suits against the AI industry.
Can training AI on copyrighted work be fair use?
Short answer: the DOJ says yes for the training step. The department separates two questions: whether ingesting millions of articles to teach a model is fair use, and whether a model’s specific output can later infringe. The brief insists the training itself is transformative and should be judged independently.
The Times sued OpenAI and Microsoft in 2023 after spotting verbatim reproductions in ChatGPT outputs.
The newspaper claims millions of articles were used without permission and that some ChatGPT responses reproduced long passages. OpenAI counters that its process converts raw text into abstracted patterns—an argument the DOJ echoed. You can see the legal battleground: publishers demand compensation; models claim a legal shield.
The administration repeatedly ties AI progress to national security and economic leadership.
President Trump has signed executive orders intended to speed AI development during his second term. The DOJ brief warns that making U.S. rules too restrictive hands advantage to foreign competitors. The framing reads like a map meant to steer judges toward a tech-friendly reading of copyright law.
Why is the US government supporting OpenAI in the Times case?
The government frames the issue as strategic: AI capabilities affect intelligence, defense, and industrial strength. When the brief says AI integration is “critical for U.S. national security capabilities,” it is asking courts to weigh broader public interests alongside private copyrights.
The Times argues for a middle path: let models develop, but pay creators for the works that power them.
The newspaper’s spokesperson told Gizmodo that creators shouldn’t be left unpaid for the content that fuels AI. I agree that sustaining journalism and creative work matters—but the question is practical: how do you bill data used to train models at scale?
The administration’s tilt favors a hands-off rule that lets firms keep training without upfront licenses; the Times wants a carve-out that preserves revenue streams for writers and reporters. This looks like two ecosystems colliding over who gets the rent.
The brief also referenced the business scale at stake, backing companies described as trillion-dollar ($1,000,000,000,000, ≈€930,000,000,000) players—names you know: OpenAI and Microsoft among them. That math helps explain the intensity: the economic stakes are enormous and immediate.
The Times frames the choice as protecting creators; the administration frames it as protecting national advantage. The dispute is part legal, part policy, and very public—but it also feels personal: Trump has long resented the paper’s coverage, and Woodward’s X post leaned into that political context.
What happens next in court and in public debate will shape AI business models.
Judges will weigh whether training is a separate, transformative act and whether outputs that copy verbatim should be treated on their own. I’ll be watching how the court treats precedent and whether it adopts the government’s separation-of-issues approach.
Two forces are pushing: one wants fast technical progress in the U.S.; the other wants to preserve payment flows to creators. The litigation could force licensing markets or confirm a broad fair-use shield—either outcome will alter product road maps at OpenAI and Microsoft, and change incentives for journalism and creative writing.
The choice before the court feels like a balance between acceleration and accountability, as sharp as a scalpel through fog. It also feels like a high-stakes poker table where lawmakers, publishers, and platforms hold different hands: OpenAI, Microsoft, The New York Times, the DOJ, and readers like you. Which hand will the judge call, and what will that call cost the public purse, the press, and the next generation of AI builders?