Unsealed Documents Reveal Microsoft and OpenAI Executives Privately Called AI Training “Theft”
aiinsider ·
Newly unsealed documents from The New York Times’ three-year copyright lawsuit against OpenAI and Microsoft expose internal admissions that executives privately viewed AI training practices as theft. Internal filings allege that the companies scraped paywalled content undetected, built massive training datasets, and deliberately removed copyright notices before feeding data to their models. Brent Hecht, Microsoft’s director of Applied Science, described the resulting decline in click-through traffic as a "doom loop" and labeled the scraping the largest theft of labor in history. Microsoft CEO Satya Nadella testified that paywalled content should be licensed prior to use and stated he would have mandated retraining had he known such material was involved. On the OpenAI side, President Greg Brockman acknowledged awareness of methods to bypass the Times' paywall, while head of ChatGPT Nick Turley noted the product's substitutive nature for publisher content. The documents claim that OpenAI’s training datasets include over 91,000 copies of materials from the Times, Daily News, and Center for Investigative Reporting. Neither company has responded to requests for comment regarding these allegations.
aiinsider