The administration of President Donald Trump has officially aligned itself with OpenAI in a high-stakes legal battle over whether artificial intelligence companies can use copyrighted works without permission to train their large language models. The administration filed a 20-page brief in the case brought by The New York Times against OpenAI, taking the position that overly restrictive interpretations of copyright law could hurt the United States’ position as a global leader in AI development.
The lawsuit is one of the most closely watched copyright disputes in the AI industry. The Times alleges that OpenAI copied millions of its articles without authorization and used them to train ChatGPT and other AI systems. In its recent brief, the administration does not explicitly decide whether OpenAI violated copyright law. Instead, it argues that the broader legal standard for fair use must not be applied in a way that would impede innovation in artificial intelligence.
Government’s core argument
The brief frames the case as not merely a dispute between two private parties, but as a matter of national importance. It argues that the United States has a strong national interest in maintaining a thriving AI sector and that American AI companies must remain competitive with rivals around the world. The administration points to an executive order signed by President Trump last year that called for retaining global leadership in artificial intelligence as an official policy objective.
“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 government brief states.
This framing suggests that the federal government is less interested in resolving the specific facts of the lawsuit and more concerned with establishing legal precedent that keeps AI companies free to train their models on large bodies of text without facing routine legal liability. The brief also warns that a ruling against OpenAI could ripple across the entire artificial intelligence ecosystem, affecting not only massive foundation model developers but also startups, researchers, and downstream products that rely on large language models.
The fair use question
At the center of the case is the legal doctrine of fair use, which allows limited unauthorized use of copyrighted material under certain circumstances. Courts generally examine four factors when deciding fair use claims: the purpose and character of the use, the nature of the original work, the amount used, and the effect on the original work’s market.
AI companies have argued that using copyrighted books, articles, and other texts to train neural networks qualifies as transformative use. Their key argument is that AI models do not reproduce the original works wholesale. Instead, the training process extracts linguistic patterns, semantic associations, and broader knowledge from millions of examples. According to this view, the resulting model is a new creation, not a substitute for the originals.
Publishers and writers reject that reasoning. The New York Times and other copyright holders argue that “training data” is simply unpaid copying on an unprecedented scale. They have also expressed concern that AI chatbots can sometimes quote, paraphrase, or summarize copyrighted works in ways that compete directly with the original publishers. Those outputs, they argue, pose a real commercial threat to newsrooms and authors who depend on subscription and royalty income.
The fair use debate in AI copyright cases is therefore not purely academic. It determines whether AI companies can continue operating their current training pipelines with relative freedom, or whether they must rebuild those pipelines through expensive licensing agreements, opt-in systems, or restricted data sets.
A legal environment increasingly favorable to AI companies
The government’s intervention comes after several high-profile lawsuits produced at least partial wins for AI companies. In one prominent case brought against Anthropic, a group of fiction writers accused the company of ingesting their books to train its Claude models. Last year, federal judge William Alsup ordered Anthropic to pay $1.5 billion in a copyright settlement. But the crucial detail in that case was narrow: Anthropic was not penalized for using copyrighted books to train its models as such. Rather, the company was punished for obtaining many of those books from illegal shadow libraries that hosted pirated copies. Judge Alsup explicitly drew a distinction between the act of reading for training and the unlawful gatekeeping used to acquire the source texts.
In his opinion, Alsup compared Anthropic’s language model training to a human writer reading published works. “Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works not to race ahead and replicate or supplant them — but to turn a hard corner and create something different,” he wrote.
That analogy has become a common talking point among AI advocates. They argue that if a human is allowed to read a novel and learn from its style, plot construction, and vocabulary, then an artificial intelligence should similarly be allowed to learn from texts without paying a separate licensing fee. Opponents respond that machine learning operates at a scale no human can match, and that ingesting millions of copyrighted books and articles is qualitatively different from a writer reading a few volumes in a library.
Other cases have produced mixed results. A class action brought by writers against OpenAI over its use of copyrighted books was partially dismissed, but the judge allowed some claims to proceed. A separate lawsuit against Meta over the training of its Llama models has also continued through the courts. Meanwhile, several visual artists have sued AI image generators such as Stability AI and Midjourney over the use of their artwork in training data sets. Some of those lawsuits have been narrowed, while others have survived initial motions to dismiss.
The broad legal picture, then, is still unsettled. But the trend so far has favored the idea that merely training AI on copyrighted material does not itself constitute infringement. The Trump administration’s brief continues that pattern at the executive level, signaling that the government may be reluctant to support legal theories that would force AI companies to pay retroactive royalties on all their past training data.
The significance and limitations of the government brief
It is important to note that the Trump administration’s brief is not a court ruling. The case is being tried in the U.S. District Court for the Southern District of New York, and the brief’s authors do not have authority to decide the outcome. Judges are not obligated to follow the views of the executive branch when interpreting copyright law.
Nor is the brief binding on other federal courts, where several similar copyright disputes are pending. Each case will ultimately hinge on its own facts, the location of the court, the specific AI system in question, and the evidence presented about how training data was collected and used.
Still, the government’s intervention carries rhetorical weight. Courts sometimes look to the executive branch for guidance on matters of national policy, particularly when a decision could have broad economic and geopolitical consequences. If this case reaches an appeals court or the Supreme Court, the Solicitor General could make similar arguments, potentially elevating the issue to the highest level of the judiciary.
The brief also adds to a growing record of federal government attention to AI. Since the release of ChatGPT, lawmakers have introduced a wide range of AI-related bills addressing subjects such as election deepfakes, algorithmic discrimination, and national security risks. But funding bills and executive orders are not the only tools shaping AI policy. Filings like this one in private litigation represent another pathway through which the government influences the rights and limits of AI companies.
Implications for publishers and the broader news industry
For The New York Times and other news organizations, the government’s decision to side with OpenAI is a significant political setback. They have argued that journalism is expensive to produce and that AI companies are effectively taking the value of that journalism without compensating the people who wrote and edited the stories. News industry trade associations have filed their own briefs in various cases, urging courts to recognize the real harms that unpaid AI training data can cause to publishers and to the public’s access to reliable information.
Some publishers have chosen an alternative path by signing licensing deals with OpenAI and other AI companies. The Associated Press, Time, News Corp, and several other outlets have reached commercial agreements that permit their stories to be used in AI training. These deals are sometimes described as a pragmatic way to create a new revenue stream amid ongoing uncertainty over the law. But not every news organization can easily strike such agreements, and smaller outlets may be left without the legal or business leverage to negotiate favorable terms.
The administration’s brief is unlikely to be the final word on whether AI companies need permission before training on copyrighted material. But it does signal that the federal government’s executive branch is prepared to defend a permissive legal environment for AI development. That stance may shift the broader public debate even as judges continue to weigh the details of individual cases.
As the New York Times lawsuit and other copyright proceedings move forward, the central question remains unresolved: whether AI companies can continue to absorb the published knowledge of humanity without paying every copyright owner whose works are used in training. The answer to that question may ultimately depend on how courts interpret the transformative purpose of large language models and how much importance they assign to American leadership in artificial intelligence.
Source: TechCrunch News