Microsoft Quietly Admitted AI Scraping Was the Largest Theft of Labor in Human History

TechCrunch (Rebecca Bellan, 17 September 2026) reported that an unredacted motion in The New York Times copyright case against OpenAI and Microsoft quotes a Microsoft director calling news scraping the largest theft of labor in human history. Ars Technica (Ashley Belanger) carried the same unsealed brief the same day.

This is an unsealed motion for summary judgment from news plaintiffs. It is not a verdict, not a Microsoft product launch, and not a regulator order. The quotes sit in the Times brief. TechCrunch notes the underlying exhibits remain sealed and the lines appear without their original context.

Microsoft Director of Applied Science Brent Hecht, in a January 2023 internal memo quoted in the filing, called scraping news for training "an astonishing theft of unprecedented proportions" and perhaps the "largest theft of labor in human history." In another document he said the plan would "make a complete mockery of the idea of fair use."

A January 2024 Hecht presentation, also quoted in the brief, said Copilot's "answer engine" cut New York Times domain click-through rates by as much as 93 percent versus traditional Bing search. Hecht framed a "doom loop" that would hurt Microsoft's models and the web at the same time. Ars Technica said Microsoft recorded 83 to 93 percent CTR drops for some plaintiffs and 51 to 94 percent for others.

OpenAI's ChatGPT head Nick Turley wrote that publishers face an "existential threat" from products that are "largely substitutive" and will get more substitutive as they improve. Coverage also quotes cofounder Greg Brockman calling the models "excellent at news." The brief alleges paywall circumvention, mass scraping, and stripping of copyright notices from training data.

A Microsoft spokesperson told Ars that Hecht's documents "reflect one employee's individual perspective" and "do not represent the company's views." The case began in 2023. Earlier this month the Trump administration filed a brief defending unlicensed training, and judges have often been friendly to fair-use arguments. None of that is a finding of theft.

Related publisher and platform tape includes the Seattle Times and Newsday complaint against the same defendants, OpenAI's new misalignment disclosure framework, and Microsoft's Mustafa Suleyman on Claude and consciousness.

If you run a newsroom or an AI training deal this week, treat the 93 percent Copilot CTR drop as the commercial fact on the table, and decide whether to license, cut the crawl, or join the Times motion before anyone waits on a trial.