Microsoft Director Called AI Scraping the Largest Theft of Labor in History. It's Now a Court Exhibit.

An unredacted court filing made public on September 17, 2026 confirms who said it: Brent Hecht, Microsoft's director of applied science, wrote in a January 2023 internal memo that AI companies scraping the web for training data amounted to "an astonishing theft of unprecedented proportions" and "the largest theft of labor in human history." The line had circulated in paraphrase for weeks. Now it has a name, a date, and a paper trail attached, courtesy of The New York Times' copyright lawsuit against Microsoft and OpenAI.

The paper trail behind one sentence

The filing, unsealed as part of a summary judgment motion, lays out how OpenAI and Microsoft acquired New York Times content well before the lawsuit was filed. According to TechCrunch, OpenAI's mid-training datasets alone hold more than 91,692 copies of articles from the Times, the Daily News, and the Center for Investigative Reporting. A Common Crawl-derived dataset pulled in more than two million documents from nytimes.com by itself, and a Microsoft-OpenAI data-sharing effort called Project Mango assembled at least 160,903 unique works from news publishers.

A 'doom loop,' documented a year later

Hecht did not let the subject drop. In a January 2024 internal presentation, he warned that Microsoft's AI content strategy had triggered a "doom loop" that would hurt the performance of its models and the entire web at the same time. The same document called the situation highly unusual: an end product competing with the economic foundations of the suppliers it depends on. Separately, CEO Satya Nadella testified under oath that clicks to news sites fell by more than 90 percent on Bing after AI summaries took over, a figure 404 Media reported lines up with the 93 percent number in Microsoft's own filing.

“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.”

OpenAI's own paper trail

OpenAI executives left a similar record. Nick Turley, who leads ChatGPT, wrote in mid-2023 that publishers faced an "existential threat" from products that are "largely substitutive" for journalism, then followed up in February 2024, per Futurism, to say the substitution problem would only grow as the models improved. OpenAI president Greg Brockman called the models "excellent at news" in an internal exchange and, according to the filing, replied "ah nice" after a researcher told him about a way to bypass the Times' paywall undetected.

Microsoft's defense: one employee, not the company

Asked about Hecht's memo, Microsoft did not disown the document, just the conclusion. "Microsoft's position is set out in its court filings, which explain why these transformative uses are consistent with copyright law and why Copilot is not a substitute for publishers' journalism," spokesperson Alex Haurek said, in a statement reported by Engadget. Neither Microsoft nor OpenAI answered a separate request for comment from TechCrunch about the filing itself. Steven Lieberman, counsel for the New York Daily News, took the opposite view in a statement to TechCrunch: "The evidence revealed here for the first time shows that OpenAI and Microsoft knew that what they were doing was wrong."

Real friction, not a slam dunk

None of this settles the underlying legal question. Several AI copyright cases have already gone the industry's way, with judges ruling that training on copyrighted material counts as transformative fair use rather than infringement, per Engadget. Those same judges have been explicit that the law here is unsettled, not resolved in AI companies' favor as a rule. The Times' case is treated as a bellwether for exactly that reason: it is the one most likely to test whether a researcher's own words, calling the product 'a product that destroys its supply chain,' can survive a fair-use defense once a jury reads them.

For a company that has spent three years telling advertisers, publishers, and regulators that generative AI creates more value than it destroys, having a senior researcher's own phrasing turn into a court exhibit is a specific kind of bad. The 91,692 copied articles are a discovery problem. The doom-loop memo is a strategy problem. The line about the largest theft of labor in human history is the one that will keep getting quoted, mostly because Microsoft wrote it first, in its own building, three years before anyone outside subpoenaed it.