© ROOT-NATION.com - Use of content is permitted with a backlink.
The Donald Trump administration has stepped into the high-profile dispute between The New York Times and OpenAI over the use of copyrighted material to train AI systems.
There’s an old legal joke: if a crime is big enough and technologically advanced enough, it suddenly stops being a crime and becomes “innovation.” That is precisely the metamorphosis the U.S. legal system is witnessing these days – and, judging by all indications, the Donald Trump administration is ready to actively help make it happen.
TABLE OF CONTENTS:
Is Trump Taking OpenAI’s Side in Its War with the NYT?
The U.S. government has effectively opposed an approach under which AI companies could be required to license virtually all of the material used to train their models. In the administration’s view, such requirements could create additional barriers for developers, limit competition in the market, and ultimately slow the development of artificial intelligence technologies.
The dispute has a much broader context than a typical conflict between a media organization and a technology company. At its core is a fundamental question: where should the line be drawn between protecting copyright and allowing AI systems to be trained on vast amounts of information that has already been published?
The New York Times and other rights holders argue that using their content without an appropriate license can cause economic harm and effectively turn someone else’s intellectual property into a free resource for commercial AI systems. Technology companies, meanwhile, argue that excessively strict licensing requirements could significantly increase the cost of developing AI models and put U.S. companies at a disadvantage compared with competitors from other countries.
The Trump administration has therefore found itself at the center of one of the key debates surrounding the future of artificial intelligence: how to protect creators’ rights, preserve competition, and avoid slowing down technological development in the United States – all at the same time.
A Dispute That Has Turned into Government Policy
The story began back in December 2023, when The New York Times filed a lawsuit against Microsoft and OpenAI. The allegation was fairly straightforward: the newspaper accused both companies of using millions of its materials – articles, reviews, and analyses – to train language models without permission and, of course, without paying a single cent in royalties. The Times demanded not only compensation but also a ban on further use of its texts, including the destruction of models and training datasets containing this content. That, it has to be said, is a rather radical demand. In practice, it would mean dismantling part of the infrastructure on which modern generative AI is built.
For nearly two years, the case quietly made its way through the courts, much like dozens of similar lawsuits against other companies. Then, on September 1, the U.S. Department of Justice decided that staying silent was no longer an option and filed what is known as a “Statement of Interest” with the U.S. District Court for the Southern District of New York. This is a document through which an entity that is not formally a party to the case explains to the court why it believes a particular position is the correct one. In this case, according to the DOJ, that position happens to be OpenAI’s.
Read also: OpenAI’s Astra: When an Algorithm Reaches What Took Mathematicians Decades to Achieve
A “Highly Transformative” Process
The administration’s argument is built around the doctrine of fair use, which under U.S. law allows copyrighted material to be used without the rights holder’s permission under certain circumstances – for example, for educational, research, or transformative purposes. Deputy Assistant Attorney General Stanley Woodward Jr. and Assistant Attorney General Brett Shumate argue that training an AI model is precisely such a transformative process. The model supposedly does not “republish” articles, but instead learns linguistic patterns and relationships so that it can later generate its own responses.
The wording is beautiful, almost poetic. It does, however, conveniently sidestep one small detail: in order to “learn linguistic patterns and relationships,” the articles first have to be copied in their entirety – and then, as the Department of Justice itself argues, used “as training material.” In other words, the copying still happens. The document simply suggests that we stop calling it copying and instead treat it as a technical step on the way to something bigger and more important.
The administration also emphasizes that if a model happens to reproduce a fragment of copyrighted text verbatim for a user, that is supposedly a different matter – a separate legal issue that should be considered independently from the question of whether the training itself was lawful. A convenient division of responsibilities: training is not infringement, while the fact that the model occasionally retransmits what was allegedly taken almost word for word is merely an unfortunate detail that can be dealt with later – and preferably somewhere else.
The Competitiveness Argument
The Justice Department’s main trump card is not so much legal as economic and geopolitical. The logic is simple: if a court rules that training LLMs on copyrighted material constitutes copyright infringement, companies will have to license the entire body of data used to train their models. And according to the administration, that would be so expensive that only the biggest players could afford such a luxury, while smaller startups and independent developers would be left out of the technology race. So, the argument goes, a restriction intended to protect creators would ultimately hurt competition and slow the development of American AI compared with foreign competitors – read: Chinese ones.
The argument is not entirely devoid of logic when considered in isolation. The problem is that it works just as well as a justification for virtually any large-scale appropriation of someone else’s property: respecting owners’ rights is always more expensive than ignoring them, while competitors are always somewhere nearby and always ready to take advantage of your hesitation. By that logic, virtually anything can be justified – the only questions are how loudly you make the argument and how influential an ally you manage to find.
Read also: How to Keep Your Child Safe Online: A Parent’s Guide
Who Benefits from the “Transformation”
The Justice Department also insists that training a model does not make the copied material “publicly available” – the texts supposedly pass through the model merely as raw material, without replacing the original in the market. This claim looks particularly ironic given how many users today turn to chatbots instead of opening a publisher’s website. But formally, of course, the “original” does not disappear because of that; it simply becomes less and less necessary.
In the final part of the document, the administration explicitly urges the court to reject the idea that training LLMs on copyrighted texts is automatically a copyright violation, calling the alternative approach – one that would require obtaining licenses before training even begins – “problematic” and legally untenable.
What This Really Means
The U.S. government’s position in this case is not merely a legal detail in a single court proceeding. It is a signal to the entire industry: the government is prepared to side with technology companies even when the issue involves large-scale use of someone else’s intellectual property, provided that use can be framed as innovation and tied to national competitiveness. The NYT v. OpenAI case has every chance of becoming a precedent that will define the rules of the game for dozens of similar lawsuits that have already piled up in U.S. courts.
And for the publishers whose archives have become the fuel feeding billion-dollar valuations for technology companies, there is little left to do but take solace in the phrase “highly transformative process” – a beautiful turn of phrase that, unfortunately, does little to compensate for lost ad impressions and readers who now get news summaries from chatbots instead of going to the original source.
Read also: Next-Generation Flying Radar: Everything You Need to Know About GlobalEye
