The Trump administration has filed a court letter supporting OpenAI in its copyright dispute with The New York Times, arguing that training artificial intelligence systems on copyrighted material may qualify as fair use.
The New York Times sued OpenAI and Microsoft in 2023, alleging that the companies violated copyright law by using its articles to train AI tools without permission. The federal government has now intervened, asserting that OpenAI’s use of copyrighted content for AI training falls within the boundaries of existing copyright law.
On Tuesday, the federal government submitted a letter telling the court that the United States has a “strong interest” in determining whether AI training on copyrighted works constitutes “fair use.” The government said the issue could affect the ability of the U.S. artificial intelligence industry to “maintain global leadership in artificial intelligence.” Fair use is a legal doctrine that permits certain uses of copyrighted material without the copyright holder’s permission.
The government argued that The New York Times’ interpretation of fair use is “inconsistent” with current copyright law. It also warned that a ruling in favor of the Times could harm the AI industry and the broader U.S. economy. “Restricting the development of LLMs based on a misinterpretation of the fair use doctrine would impede such creative and scientific progress and, at the same time, impede American prosperity and economic mobility,” government lawyers wrote.
Although the letter centers on the lawsuit involving The New York Times and OpenAI, the government said its legal arguments apply to all parties in the case and related litigation, including publishers, authors, and other AI companies.
The government described training large language models, or LLMs, on copyrighted works as “highly transformative” because AI systems process source material to produce new and meaningful outputs. It also argued that OpenAI’s models do not directly compete with individual New York Times articles. Courts will weigh these and other factors when deciding whether AI training meets the requirements of the fair use doctrine.
The letter also cautioned against equating the process of training an AI model with the content produced by an AI tool. Government lawyers said that doing so could imply that human writers violate copyright law simply by studying and practicing with other authors’ work. They compared the issue to reports that a young Joan Didion typed out Ernest Hemingway stories as a way to learn the craft of writing, warning that a broader interpretation could have “troublesome implications for copyright law generally.”
Evan Brown, an intellectual property attorney with Neal & McDevitt, told WIRED that U.S. District Judge Sidney H. Stein, who is overseeing the case, is not required to follow the government’s position. However, Brown said Judge Stein and other judges handling AI copyright lawsuits will likely “take this matter very seriously because it comes from the Department of Justice and inherently carries a lot of weight.”
Pamela Samuelson, a copyright lawyer and co-director of the Berkeley Law and Technology Center, called the government’s filing an “important development” that is consistent with previous court decisions involving artificial intelligence and copyright law.
“The administration is siding with a handful of trillion-dollar AI companies at the expense of the countless American creators whose work has been stolen,” The New York Times spokesperson Graham James told WIRED. “AI and creators can both thrive. AI companies just need to follow copyright law and pay fairly for the content that makes their products possible.”
OpenAI and Microsoft did not immediately respond to WIRED’s requests for comment.
Other copyright plaintiffs also criticized the government’s intervention. Mary Rasenberger, chief executive of the Authors Guild, which filed its own lawsuit against OpenAI in 2023, said the organization was “extremely disappointed” with the letter. She described it as containing “false arguments and serious misinterpretations of fair use doctrine and copyright law.”
Dozens of lawsuits continue to challenge whether AI companies can legally use copyrighted books, articles, and other creative works to train their models without permission. Two of the most significant rulings so far came last year. In Kadrey v. Meta, Meta technically prevailed in a copyright lawsuit, but the judge said the plaintiffs had not provided enough evidence that AI training had harmed them. The court also emphasized that using copyrighted material to train AI systems without permission could still be unlawful under different circumstances.
Source: www.wired.com