Trump Administration Backs OpenAI in Copyright Fight Over AI Training

The Trump administration has filed a brief supporting OpenAI in its battle with The New York Times over whether AI companies may train large language models on copyrighted material without permission. The filing argues that protecting the United States’ AI sector is a national priority and that an overly narrow reading of copyright law could slow innovation.

The government’s intervention does not decide the case, but it adds political weight to one of the most consequential legal fights in artificial intelligence: whether using books, articles and other protected works to train models such as ChatGPT is lawful fair use or copyright infringement.

What the government said in its filing

The administration’s 20-page brief takes OpenAI’s side on the broader policy question at the heart of the lawsuit. According to the filing, the United States has a strong interest in keeping domestic AI development competitive and maintaining global leadership in the technology.

In practical terms, the brief argues that courts should be careful not to interpret fair use so narrowly that the country’s AI industry loses momentum. The government says that constraining model training too aggressively could hurt scientific progress, creativity, economic mobility and American prosperity.

The brief says that preserving a robust AI ecosystem is essential for the United States to remain a global leader in artificial intelligence, and warns that a restrictive view of fair use could slow innovation across the economy.

This is not the same as a ruling. The federal government is not the judge in the case, and its filing does not settle whether OpenAI’s training methods were lawful. But in a high-stakes copyright dispute, the position of the White House can still influence how courts, policymakers and the public think about the issue.

Why is training on copyrighted material such a big legal issue?

Because modern AI systems are built by ingesting enormous datasets that often include books, news stories, web pages, images and other works protected by copyright. Companies typically do not seek individual licenses for every item included in those datasets, which has led publishers, authors and media organizations to accuse AI firms of using their work without consent.

The legal debate centers on whether that use is covered by fair use, the copyright doctrine that allows limited unlicensed use of protected material in specific circumstances. The key question is whether training a model on copyrighted works is sufficiently transformative to qualify as lawful under U.S. law.

AI companies generally argue that model training does not copy works for direct consumption in the way a book seller or media distributor would. Instead, they say the training process helps the system learn patterns and relationships so it can generate new output. Critics counter that the companies are still exploiting protected works at scale, often without compensation.

How does fair use apply to AI training?

Fair use can protect certain uses of copyrighted material when the new use changes the original work enough, serves a different purpose, or advances scholarship, commentary or research. In the AI context, companies argue that training is transformative because the model does not reproduce the source material as a substitute for it.

Rights holders disagree, saying the output of these systems can compete with the original works and that the training process itself may involve unauthorized copying. Courts are now being asked to decide how much transformation is enough and whether the scale of AI training changes the analysis.

How this case fits into the wider AI copyright fight

This lawsuit is part of a broader wave of cases testing how copyright law applies to generative AI. News publishers, book authors, artists and record labels have all raised similar complaints: that companies trained models on their work without asking first.

Some early decisions have leaned toward the AI companies, at least on the training question. But the legal picture remains unsettled, and judges are still sorting out whether the training process, the data sources or the model outputs are the legally important part of the dispute.

That uncertainty matters because the outcome could shape how AI companies build their systems for years to come. A ruling that requires broad licensing could raise the cost of training and alter the economics of model development. A ruling that favors fair use could accelerate AI deployment and weaken the leverage of publishers seeking compensation.

What happened in the Anthropic settlement?

One major reference point is the Anthropic case, which produced a large settlement but not a broad condemnation of model training itself. A federal judge ordered the company to pay $1.5 billion to a group of writers after their books were found to have been sourced through unauthorized shadow libraries, but the court did not punish Anthropic simply for training on books in the abstract.

That distinction matters. The case suggested that the way the data was obtained can be just as important as the fact that it was used. In other words, courts may be more willing to scrutinize piracy and illegal acquisition than the mere act of training an LLM on copyrighted material.

In that earlier decision, the judge likened model training to a person reading in order to learn, while still drawing a line at the use of pirated source material.

For AI companies, that is encouraging but not a blank check. It hints that the source of training data, the surrounding practices and the output behavior of the model all matter to the legal analysis.

Why does the Trump administration care about OpenAI’s case?

The administration’s position reflects a broader competition policy: the United States wants to dominate AI development rather than cede influence to other countries. By framing AI leadership as a matter of national interest, the brief places copyright law in a strategic context.

That approach matters because governments increasingly see AI as an infrastructure issue, not just a consumer technology question. Models are becoming embedded in search, office software, customer service, coding tools and media workflows. If American firms face stricter legal barriers than foreign rivals, policymakers worry the U.S. could lose ground.

The brief also echoes a recurring argument from the tech industry: that AI training should be treated as a foundational innovation process, similar to reading, indexing or statistical analysis, rather than as a simple act of reproduction. Supporters say that legal flexibility is necessary if the United States wants to retain its edge in a fast-moving field.

What does the filing actually change?

It does not change the law on its own. But it can shape how judges interpret the stakes, and it signals that the federal government sees the issue as bigger than one company’s lawsuit.

Government briefs can also be useful to companies in similar disputes because they help establish a policy narrative. If a court is undecided on a close fair use question, a filing from the executive branch may provide additional context on how a ruling could affect national competitiveness, innovation and public policy.

Timeline of the key developments

Below is a simplified timeline of the major legal and policy moments referenced in this dispute.

What does this mean for publishers and creators?

For publishers, the administration’s filing is a setback in the broader campaign to force AI companies to pay for training data. Media organizations argue that their reporting, books and archives are being repurposed into commercial products without consent or compensation.

Creators also worry about market substitution. If a chatbot can answer questions by summarizing or rephrasing articles, books or research, the original source may lose traffic, subscriptions or sales. That concern is one reason many rights holders want courts to draw a clearer legal line around model training.

Still, the legal fight is not just about money. It is also about control, attribution and the future rules of digital creativity. A favorable ruling for AI firms could normalize broad data scraping for training. A ruling for publishers could force the industry toward licensing deals and tighter dataset governance.

• Publishers want compensation and consent for use of their work.
• AI firms want enough legal room to train models at scale.
• Courts are being asked to decide whether training is transformative enough for fair use.
• Government support for AI companies strengthens the policy case for permissive training rules.

How are courts likely to think about this next?

Courts are likely to keep focusing on two linked questions: where the data came from and what the model does with it. Judges may be more skeptical of datasets built from pirated or illegally obtained material, while remaining open to fair use arguments for lawful copies used in a transformative way.

They will also likely examine whether an AI system reproduces protected material too closely, or whether it simply learns from it in a way that resembles human reading and synthesis. That distinction has become one of the defining issues in generative AI litigation.

Why the outcome could shape the AI industry

The outcome could determine whether frontier AI companies need to license much of the material that underpins their systems. That would affect cost, model development timelines and bargaining power between tech firms and rights holders.

It could also influence where AI research happens. If U.S. law becomes too restrictive, companies and labs might push more work into jurisdictions with looser rules. That is one reason policymakers are treating these cases as part of a larger industrial strategy.

At the same time, a permissive ruling could deepen the conflict with creators who say their work is being used to build products that compete directly against them. The court’s decision may therefore do more than settle a copyright question — it may define the balance of power between the AI industry and the creative economy.

What happens next?

The New York Times case will continue in the U.S. District Court for the Southern District of New York, where the judge will weigh the claims, defenses and broader legal record. The Trump administration’s brief is now part of that background, but it is not the final word.

Expect more briefs, more expert arguments and more attention on how the court interprets fair use in the age of large language models. However the case ends, it is likely to become one of the most closely watched copyright decisions in the AI era.

For now, the clearest takeaway is that Washington is signaling it wants the U.S. AI industry to remain aggressive, competitive and legally viable — even as publishers and creators continue pressing for stronger protections over the material that helps train those models.