AI Training Called "Largest Theft In Human History" By Microsoft Exec In Court

You might prefer one of the Microsoft exec's other quotes during a deposition in a lawsuit about whether AI companies taking data and other content for AI training without permission or payment is 'fair use' or stealing: it is "an astonishing theft of unprecedented proportions."

Either way, the word 'theft' is central to the argument that AI companies are taking something that is not theirs and building a trillion dollar business model from it. That valuation depends, in large measure, on the fact that the AI business has almost zero cost of goods sold because the content on which their products and services are built has been stolen. That this quote comes from tech execs underscores the economic and ethical questions that have arisen inside the industry, but that the companies have attempted to shield from public view. Should they actually be forced to pay for the intellectual property they are using for free, it would significantly reduce their margins and further undermine the content and profit grab which has, so far, powered AI and its financial projections. JL

Scott Nover and Gerrit De Vynck report in the Washington Post:

A Microsoft executive called the training of AI models the “largest theft of labor in human history,” according to a legal brief. That same executive called it “an astonishing theft of unprecedented proportions.” The material unearthed in discovery shows how tech titans have built a business model worth billions off the backs of news organizations, whose work has been fed into AI tools in vast quantities to “train” them, often without permission. Using this material, the AI giants have produced a behemoth that conveys information, without properly crediting — or paying for — the original work. Novelists, musicians and other creative professionals argue the companies stole their work and used it to build tools that can be used to replace them. For the plaintiffs, Microsoft’s admission amounts to evidence that Microsoft and OpenAI knew they were stealing journalism without permission or payment.

A Microsoft executive called the training of artificial intelligence models the “largest theft of labor in human history,” according to a legal brief filed by the New York Times and unsealed Thursday. In another instance, according to the brief, that same executive called it “an astonishing theft of unprecedented proportions.” 

The comments come from Brent Hecht, Microsoft’s director of applied science, and headline a legal brief filed by the New York Times and other news publications in their high-profile copyright case against the tech giant Microsoft and the AI company OpenAI. Hecht is also a professor at Northwestern University. 

The material unearthed in discovery, the plaintiffs argue, shows how the two tech titans have built a business model worth billions off the backs of news organizations, whose work has been fed into AI tools in vast quantities to “train” them, often without explicit permission from the outlets. Using this material, the plaintiffs argue, the AI giants have produced a behemoth that conveys information to readers, without properly crediting — or paying for — the original work of journalists.

Hecht’s comments “reflect one employee’s individual perspective, are not a legal analysis, and do not represent the company’s views,” a Microsoft spokesperson said.

AI executives have argued that their training material falls under the legal principle of fair use, and that they have not improperly used or delivered news content. But the brief, which references emails and company records obtained during discovery, as well as depositions with key executives, highlights moments that show internal discord over the practice of training AI models on journalism, revealing a tension between the companies’ public legal arguments and private debates over the ethics of the practice.

The documents also surface moments where AI leaders showcase the technology as adept at replicating journalism. It “seems to be particularly good at predicting text of news articles,” wrote OpenAI president and co-founder Greg Brockman in a 2020 message. “Like whenever i have it complete in the middle of a sentence in a NYT article, it seems to complete the sentence on point.” Brockman reiterated the point, according to the brief, writing that AI models are “very good at any news task.” 

In a deposition, Microsoft CEO Satya Nadella said that chatbots provide “information right there on the website on the AI platform versus needing to go to the underlying source,” according to the brief.

Nadella’s “testimony and Microsoft’s position in this case are perfectly consistent. He spoke to broad principles and changes underway in how people find and consume information,” the Microsoft spokesperson said.

The Times is asking for a federal judge to grant summary judgment finding the two technology companies liable for “unauthorized copying at each stage of the AI pipeline.”

For the Times and its fellow plaintiffs, Microsoft’s admission amounts to a smoking gun — evidence that Microsoft and OpenAI knew that they were stealing journalism without permission or payment.

“The evidence revealed here for the first time shows that OpenAI and Microsoft knew that what they were doing was wrong,” said Steven Lieberman, a lawyer for the news companies. “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. Finally, the world can see what OpenAI and Microsoft thought all along about the fairness of their own behavior.” 

Spokespeople from OpenAI did not respond to a request for comment. (The Washington Post has a content partnership with OpenAI.)

Modern AI models are trained on vast amounts of data scraped from the open web. AI companies have contended that using publicly available information to train their AIs is similar to a student reading books to learn about the world. Novelists, musicians and a host of other creative professionals have pushed back, arguing that the companies stole their work and used it to build tools that can be used to replace them. 

The Times and fellow plaintiffs argue the evidence presented in the brief shows that OpenAI and Microsoft knew that their tech could be used to substitute news services.

The Times first sued Microsoft and OpenAI in 2023, about one year after the latter released ChatGPT. It was joined by a slew of other news organizations including the New York Daily News, the Intercept, and the nonprofit Center for Investigative Reporting.

Microsoft and OpenAI have claimed that their actions are covered by the legal doctrine of fair use, a set of allowances for using copyrighted works.

In its brief, the Times said that after it sued OpenAI, the company set up a filter to limit any output that could subsequently be used as evidence against them, specifically targeting the plaintiffs.

It’s not the only high-profile case pitting legacy publishers against AI juggernauts: The Times and the Wall Street Journal are each suing the AI search engine Perplexity. Anthropic, the maker of Claude, also settled with book authors for $1.5 billion in 2025.

During a June 1 speech, Times Publisher A.G. Sulzberger accused AI companies of “hijacking of the public square” by stealing journalism. “They repackage these stolen goods as their own, siphoning off the audiences and revenue that otherwise would go to the news organizations that created this work,” he said at an event.

This poses a threat to journalism and the public, he said: “A future where a crucial wellspring of a healthy society and a stable democracy — the truth, understanding and accountability provided by original journalism — continues to dry up.”