Court document: OpenAI accused of 'the greatest theft of human effort' to train its artificial intelligence

The New York Times reported in a court document unsealed Thursday (September 18, 2026) that OpenAI committed a “stunning theft of unprecedented scale” when it used millions of news articles to train its artificial intelligence models.

She said that the San Francisco-based Artificial Intelligence Laboratory automatically collected the content of more than 10 million articles, about a third of which came from the New York Times alone.

The document quoted Microsoft's director of applied sciences, Brent Hecht, describing it as "an astonishing theft of an unprecedented size," and perhaps "the largest theft of human effort in the history of humanity." Hecht warned that OpenAI may have engaged in an "inadvertent cover-up" in its attempts to identify content on its systems that was taken from the New York Times and the other complainants in the lawsuit.

But Microsoft considered that Hecht's statements express "the individual point of view of an employee" and "do not represent the company's point of view," according to what a Microsoft spokesman told AFP.

The newspaper filed a lawsuit against OpenAI and Microsoft three years ago in a federal court in New York, accusing them of stealing its copyrighted materials to train OpenAI's main model, "Chat GPT."

Microsoft invested in OpenAI for the first time in 2019.

A number of other news publishers joined the lawsuit, including Ziff Davis, which owns CNET and other technology news publications, in addition to the parent company of Mother Jones magazine, the investigative news site The Intercept, and a number of local newspapers across the United States. They are demanding compensation for every article suspected of being stolen and used in OpenAI models, while the total amount of potential compensation remains unclear.

Since launching ChatGPT in late 2022, OpenAI has signed content licensing agreements with many news publishers around the world, but concerns remain that generative AI software is reducing its reach on online news sites.

AI companies have tried to allay these concerns by adding referrals and links in their answers to user queries, but these efforts have been criticized as insufficient. “No matter how much we highlight the links, users will not click on them,” an OpenAI engineer said, according to the court document.

Microsoft and OpenAI say their use of news content is transformative and falls within "fair use" laws.

In early September, the US Department of Justice submitted a memorandum to the court in support of OpenAI and Microsoft, citing “scientific progress,” economic growth, and “national security.” The plaintiffs requested summary judgment in their favor. If Federal Judge Sidney Stein approves the request, the case will not be referred to court, although a ruling is not expected before 2027.