Federal lawyers argue that training AI models on news content does not violate copyright law. The New York Times and other publishers accuse OpenAI of using their work without permission. The case could set new rules for how AI and publishers interact.
The Trump administration has come out in support of AI companies, telling a federal judge that using news articles to train large language models does not break copyright law. This position, laid out in a recent court filing, puts the Department of Justice on the side of OpenAI and other tech firms, and against publishers like The New York Times Co., who say their work is being used without payment.
The DOJ argued that training AI models on copyrighted material is "transformative"-a legal term that can protect AI companies from copyright claims. The government says that using articles to build an interactive model is fundamentally different from publishing or entertaining, which is the original purpose of the work. DOJ officials also said that this process does not harm the market for the original content, since the training itself is not public. According to The New York Times, the DOJ filed its statement in the U.S. District Court for the Southern District of New York on September 1-2, 2026.
This is the first known instance where the U.S. federal government has formally intervened in an AI copyright dispute on the side of AI developers, marking a significant precedent in the ongoing legal battles between tech firms and publishers.
The lawsuit, led by The New York Times Co. and other publishers, claims OpenAI violated their rights by using news stories to train ChatGPT and by reproducing parts of those stories in chatbot responses. The government’s filing, submitted to Judge Sidney Stein, echoes a White House policy from March that favors innovation and free expression over strict copyright enforcement in AI development. As The Hollywood Reporter notes, Judge Stein is handling a combined case with claims from several publishers and authors against OpenAI and Microsoft, with a deadline set for summary judgment motions.
Publishers see the government’s stance as a threat to their business. A spokesperson for The New York Times accused the administration of siding with trillion-dollar AI companies at the expense of American creators, arguing that AI firms should pay for the content that powers their products. The Times warned that letting AI companies use content without permission or payment could undermine the future of journalism. The DOJ, in its filing, responded that the "creative possibilities and public benefits" of AI model training outweigh possible harm to publishers, and warned that restricting such training could slow "creative and scientific progress" (US News).
The law on this issue is unsettled. In California, Judge William Alsup found that Anthropic’s use of purchased books to train its Claude chatbot was "exceedingly transformative" and counted as fair use. But Alsup also ruled against Anthropic for downloading millions of pirated books, which led to a $1.5 billion settlement with authors. In another case, Judge Vince Chhabria suggested that most uses of copyrighted material for generative AI training would likely be illegal, warning that such technology could flood the market with derivative works. The Justice Department criticized Chhabria’s view, saying it misapplied copyright law.
The DOJ's brief emphasized that training large language models on copyrighted materials does not make those materials publicly available nor does it directly harm the market for the original works. The department also cited national security, innovation, and economic growth as key reasons for supporting AI development, arguing that overly restrictive copyright interpretations could hinder the United States' technological leadership.
Department of Justice
For publishers and content creators, the stakes are immediate. If the courts accept the government’s argument, AI companies could keep scraping and processing huge archives of journalism without paying for it. That would shift more power to tech platforms, putting more pressure on newsrooms and threatening the business model of original reporting. The government’s claim that training does not harm the market for original works overlooks the fact that AI-generated outputs can replace the content they use, especially as models improve.
With the legal situation unclear, publishers are rethinking their approach. Some are trying to block AI crawlers with technical tools, others are negotiating licensing deals or joining lawsuits. The outcome of this case will set a precedent that could either protect the value of original journalism or speed up its commodification by AI. As reported earlier, the overlap between AI and media is already changing how content is created, checked, and sold.
By siding with AI firms, the Trump administration is making a clear choice: it believes the economic and scientific benefits of open AI development outweigh the risks to publishers and creators. But this is not a neutral move-it shifts value from those who produce original work to those who build the tools that reuse it. For media companies, the message is clear: unless courts or Congress step in, the rules for dealing with AI are changing fast, and doing nothing could have serious consequences.
