OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI and Microsoft executives internally warned their AI strategy would create a "doom loop" that damages model performance and the entire web, per unsealed court documents in the NYT case.
- Brent Hecht, Microsoft's Director of Applied Science, called the data scraping "the largest theft of labor in human history" and said the fair use defense makes "a complete mockery" of fair use doctrine.
- Satya Nadella acknowledged chatbots have replaced search, removing the need to visit source websites, while later stating "anything that is paywalled should be licensed."
- OpenAI employees admitted internally that GPT-4 "memorized a ton of data and therefore will be insanely good at regurgitation," despite acknowledging that preventing memorization was important to "minimize copyright violations."
- OpenAI representatives said they were "unaware" of any effort to detect or remove paywalled content from training data, contradicting Nadella's licensing statement.
- OpenAI's own media and economics experts attributed referral traffic drops of up to 60% for sites like the NYT to AI summaries, and cofounder Greg Brockman was focused on the "gazillions" of dollars in potential commercial AI revenue.
Why it matters: These internal admissions — captured in a 92-page filing — directly undermine Microsoft and OpenAI's legal defense that AI training doesn't harm publishers, strengthening the NYT's copyright case at trial. For the publishing industry, the documented awareness of a "doom loop" reframes the litigation from accidental infringement to knowingly destructive behavior, potentially exposing both companies to broader damages or mandatory licensing.
Ask SkimNews
