The US Department of Justice filed a statement in the class-action lawsuit brought by The New York Times against OpenAI, arguing that training artificial intelligence models on copyrighted text constitutes fair use under copyright law. This position directly opposes a recent report from the US Copyright Office, which had warned against broad fair use claims in AI training contexts.
The DOJ's filing carries enormous weight. Courts typically treat government legal positions seriously, and this statement could shape how judges interpret fair use doctrine in the AI era. Fair use allows limited copying of copyrighted material without permission under specific circumstances. The doctrine traditionally applies to parody, criticism, education, and news reporting. The DOJ argues AI training fits within these boundaries because models learn patterns rather than reproduce the original texts verbatim.
The timing reveals sharp political divisions over AI policy. The Copyright Office director was dismissed by the Trump administration shortly after publishing a report that contradicted the DOJ's current stance. That earlier report suggested fair use might not protect large-scale training on copyrighted works without licensing deals. The sudden reversal reflects a pro-AI industry stance that prioritizes technological advancement over creator protections.
The New York Times lawsuit targets OpenAI and Microsoft, seeking damages for unauthorized use of millions of copyrighted articles in training ChatGPT and related models. The case tests whether AI companies must license content or negotiate with publishers before training systems on their work. News organizations argue their copyrighted material represents their primary product and that AI training undermines their business models. OpenAI has countered that fair use allows training on internet-available text.
Federal courts have never definitively ruled on this specific question. Earlier cases involving Google Books and search engine caching offered some precedent for transformative uses, but AI training operates at a scale and scope those decisions didn't contemplate. The outcome affects whether AI companies must establish licensing frameworks with content creators or whether they can treat the internet as a training commons.
The DOJ argument emphasizes that AI models don't function as substitutes for copyrighted texts. Users cannot extract the original articles from ChatGPT. Instead, models extract abstract patterns used to generate new outputs. This transformative quality, the DOJ contends, mirrors fair use cases where copying occurred but the end product served a different purpose than the original.
Industry observers note the DOJ position signals the Biden-to-Trump transition in AI policy. The Biden administration investigated AI companies and explored stricter regulation. The Trump administration has signaled stronger support for AI industry interests, and this filing reflects that shift. The reversal could embolden other AI companies facing similar litigation.
Publishers and creators argue this interpretation would decimate licensing markets for professional content. If AI training automatically qualifies as fair use, media companies lose leverage to negotiate compensation. Journalists and authors would subsidize AI development through unpaid use of their work.
The case will likely proceed to summary judgment or trial within the next year. Whatever the court decides will ripple across AI development worldwide, determining whether AI companies can operate as quasi-utilities with free access to human-created works or whether they must compensate creators at scale.