Court filings unsealed on September 17 in a Manhattan federal court show Microsoft and OpenAI executives privately described their AI training as built on other people’s work. The documents come from The New York Times’ copyright lawsuit against the two companies.
Brent Hecht, Microsoft’s director of applied science, told colleagues that AI companies’ mass scraping of online content was “an astonishing theft of unprecedented proportions.” In another document he called it potentially “the largest theft of labor in human history.” He warned the practice could collapse the web the models depend on, calling it a “doom loop.”
The filings say OpenAI’s training sets held more than 91,692 copies of articles from the Times, the Daily News and the Center for Investigative Reporting. A separate Microsoft dataset drew on about 160,903 works from news publishers.
Microsoft’s own data showed its Copilot answer engine cut clicks to Times stories by as much as 93 percent compared with normal search. OpenAI’s head of ChatGPT, Nick Turley, wrote that such products are “largely substitutive.” OpenAI co-founder Greg Brockman allegedly replied “ah nice” when told about a way to bypass the Times’ paywall. The companies have said in public that building ChatGPT was legal.