Microsoft Exec Calls AI Training 'Largest Theft of Labor in History'

A Microsoft executive privately described AI training on scraped web content as 'the largest theft of labor in history,' according to an unsealed New York Times court filing. The remark came from a January 2023 memo by Brent Hecht, Microsoft's director of applied science, who also called the practice 'an astonishing theft of unprecedented proportions.' Hecht's writings warned that the companies' AI systems were creating a 'doom loop,' suppressing traffic to the publishers whose work trains the models and starving the content supply those models depend on.

Microsoft's own data showed its Copilot answer engine cutting click-throughs to The New York Times' domain by 83-93 percent compared with traditional Bing search. The filing details other admissions, including Nick Turley's description of commercial AI products trained on news as an 'existential threat' to publishers.

The companies argue that training on copyrighted text is transformative and protected by fair use, but the Times alleges they scraped paywalled articles, assembled training datasets containing tens of thousands of copies of the plaintiffs' work, and stripped copyright notices before the content reached the models. The case has significant implications for the future of AI development and its impact on publishers.