Microsoft disclaims its own director’s “astonishing theft”

Microsoft’s own data recorded 83–93% drops in click-through rates for some of the news organizations suing it and OpenAI, and 51–94% drops for others. The figures appear in a motion for summary judgment from the news plaintiffs led by The New York Times, unsealed on September 17. The motion also quotes internal documents in which staff at both companies called AI training on news theft and described chatbots as substitutes for the outlets they draw on.

The defendants spent years fighting to keep these documents out of public view, Ars Technica reported. The motion sets the click-through figures next to a Microsoft document warning of a “doom loop” that “will hurt the performance of our models and the entire web at the same time.” The same document says: “It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its ‘content supply chain.’” The news groups say Microsoft also included a cartoon illustrating the problem of LLMs destroying their own supply chains.

Brent Hecht, Microsoft’s Director of Applied Science, is quoted more than anyone else. The news groups say he wrote more than once that scraping news for training was “an astonishing theft of unprecedented proportions” and possibly the “largest theft of labor in human history.” In another document he wrote that the plan made “a complete mockery of the idea of ‘fair use,’” and fair use is the defense both companies rely on in court. He also wrote that “almost no one intended for content they created to be used in this fashion, nor are they compensated for its use.”

“No good reason to click”

At OpenAI, the motion cites ChatGPT head Nick Turley warning in an internal message that publishers faced an “existential threat” from commercial products trained on their content that could stand in for them. Turley also called chatbots “largely substitutive, period” and predicted they “will get more and more substitutive as they get better.” The plaintiffs say he agreed there is “no good reason to click” once the chatbot has given the user the information. An OpenAI software engineer wrote that “no matter how prominently we show the links, users won’t click.”

Satya Nadella said much the same under oath. The Microsoft CEO described the chatbot as “giving you the information right there on the website on the AI platform versus needing to go to the underlying source.” He also testified that AI companies should not break news sites’ terms of use by getting around paywalls. Inside OpenAI, staffer Nick Ryder told president Greg Brockman that “a hack” had been found for OpenAI’s crawlers “to get around” the NYT paywall. Brockman replied, “Ah, nice.” Elsewhere in the motion, Brockman is quoted saying he was “deeply motivated by the gazillions” that commercializing OpenAI’s technology could bring.

How the verbatim outputs were found

Early in the case, plaintiffs pulled articles out of chatbots by asking “what’s the next line?” over and over. The motion shows they went further. Asking for a summary sometimes produced long excerpts, and so did asking for an article’s key bullet points. Asking the chatbot to “rate the bias” of an article worked particularly well. Asking it to pick any article from a site’s homepage worked too. For now the plaintiffs ask the court to rule only on articles where outputs “demonstrate extensive verbatim overlap,” and say they will raise the other articles at trial.

The plaintiffs also say the companies did not stop such outputs and instead built a filter that made them harder to test. In the plaintiffs’ account, Hecht suggested the filter could be seen as an “accidental cover up,” because it would leave “people who have a right over the content having less visibility into what was used for training.”

Two allegations concern where the training data came from. The motion says Microsoft broke “industry norms” by selling a dataset it had bought for Bing to OpenAI as training data, without consulting publishers who had only agreed to ordinary search crawling. It also says OpenAI got a dataset of 1.8 million NYT articles from a third party that was bound by an agreement not to use the data commercially. The news groups say OpenAI employees knew it “would not be appropriate” to use that data “to train a model” and did so anyway.

Where the case stands

The Times filed the suit in December 2023 in the Southern District of New York, before Judge Sidney H. Stein. In 2025 Stein mostly denied the motion to dismiss. He did not decide whether training was fair use and said the question needed a fuller factual record. That record is now in front of him: the news plaintiffs, OpenAI and Microsoft filed cross-motions for summary judgment on fair use on September 4, and replies are due November 6. Shortly before that, the Justice Department filed a statement of interest backing OpenAI and Microsoft, reportedly the first time it had weighed in on an AI copyright case. Statutory damages for willful infringement can reach $150,000 per work, and the Times says it is seeking “billions of dollars.”

The plaintiffs’ argument rests on substitution. “With respect to outputs that are substantially similar to training or grounding sources, courts have rejected claims that copying news articles to provide a product that substitutes for demand for news is fair use,” they wrote. They also describe the situation as a prisoners’ dilemma: each AI company is better off taking content for free while others pay. Their example of another firm following the same path is Google’s AI Overviews, launched after ChatGPT. In their words, finding that copying news for AI is not fair use “would solve this prisoners’ dilemma by putting all AI companies, OpenAI and Microsoft included, on an even footing.”

A Microsoft spokesperson said the company’s AI products are transformative fair use and do not substitute for news sites. According to the spokesperson, Nadella’s testimony was a set of “observations” about “broad principles and changes underway in how people find and consume information” that “should not be confused with conclusions about copyright questions before the Court.” The spokesperson said Hecht’s documents “reflect one employee’s individual perspective, are not a legal analysis, and do not represent the company’s views.”

Steven Lieberman, counsel for the New York Daily News and seven of its sister papers, said the evidence “shows that OpenAI and Microsoft knew that what they were doing was wrong.” He added: “Throughout this case Defendants insisted that these documents be treated as confidential so that the public could not see them. Well, now the cat is out of the bag.”