Court filings made public by The New York Times include remarks from a Microsoft executive who called the use of news articles to train AI models "the largest theft of labor in human history." The documents were released in connection with lawsuits filed at the end of 2023 by The New York Times and other publishers, including the Daily News, against OpenAI and Microsoft.
The plaintiffs allege that OpenAI and Microsoft used vast amounts of news content to train language models often without obtaining publishers’ permission. According to the filings, the technology companies built lucrative products that rely on other organizations’ journalistic work while delivering information in ways that previously required consulting the original news articles.
What do the filings show?
The documents indicate that people inside the companies were aware of how well the models could mimic journalistic writing. The filings quote Greg Brockman, OpenAI president and cofounder, in a 2020 message saying the technology is "especially good at predicting text of news articles" and could finish a sentence cut off mid-article in The New York Times. Brockman later wrote that the models are "very good at any tasks related to news."
The filings also include testimony from Satya Nadella, chief executive officer of Microsoft, who said chatbots provide information "in-platform, in the moment, rather than a person having to go to the original source."
The New York Times and other plaintiffs say these documents show Microsoft and OpenAI knew they were using journalists’ content without permission or payment, and they are asking a federal court to find the two companies liable for unauthorized use during AI training.
Responses and legal arguments
Microsoft has tried to distance itself from the quoted remarks. A company spokesperson said the comments attributed to Brent Hecht reflect "one employee’s personal view, are not a legal analysis, and do not represent the company’s position."
Both Microsoft and OpenAI have previously argued that fair use law can apply to the training of artificial intelligence—an argument that, if accepted, would allow some use of copyrighted material without permission under certain conditions.
Steven Lieberman, the attorneys’ lead for the publishers, said the newly surfaced filings cast the companies’ conduct in a different light. "The evidence revealed here for the first time shows that OpenAI and Microsoft knew what they were doing was wrong," he stated, and he noted that defendants had sought to keep these documents confidential.
Why the case matters
The lawsuits go beyond individual damages and could set precedent for how much copyrighted journalistic material large language models may lawfully use for training without permission or payment. The internal remarks and testimony disclosed in the filings may play a significant role as the federal court considers liability for the alleged unauthorized use.
Lead photo for illustration only. Photo: AFP / Jason Redmond
