Unsealed NYT v. OpenAI filings show Microsoft and OpenAI executives called AI scraping 'the largest theft of labor in human history'

Newly unredacted summary-judgment filings in the New York Times v. OpenAI and Microsoft copyright case reveal internal admissions that LLMs were trained on copyrighted works via paywall circumvention, that executives viewed chatbots as 'largely substitutive'…

TechCrunch, Ars Technica and 404 Media (all 2026-09-17) report on newly unsealed material in the three-year-old New York Times-led copyright suit against OpenAI and Microsoft. A January 2023 internal memo by Microsoft's Brent Hecht called AI scraping 'an astonishing theft of unprecedented proportions' and 'the largest theft of labor in human history,' while OpenAI's Nick Turley described chatbots as 'largely substitutive' and an existential threat to publishers. A Microsoft document also warned its AI content strategy started a 'doom loop' threatening the web's content supply chain. On traffic impact, sources frame the figures differently: TechCrunch says Microsoft's Copilot answer engine cut NYT click-through rates by up to 93% versus Bing; Ars Technica cites Microsoft-recorded 83-93% click-through-rate drops for some news plaintiffs without the Bing comparison; and 404 Media reports Satya Nadella testified under oath that Bing referrals to news sites fell by more than 90% after content was used for AI answers. OpenAI datasets reportedly contained 91,692 copies of the plaintiffs' works and over 2 million nytimes.com documents, with Project Mango holding 160,903 unique works. Internal messages show a crawler 'hack' that let OpenAI bypass the NYT paywall — which Greg Brockman applauded ('ah, nice') — and employees allegedly stripped copyright notices from training data. Nadella testified that paywalled content should be licensed for grounding or training, and plaintiffs are limiting summary judgment to articles with extensive verbatim output overlap. The admissions cut against OpenAI's fair-use defense, which the Trump administration recently supported in a brief.

• An unredacted summary-judgment motion was unsealed in New York Times v. OpenAI and Microsoft, a three-year-old copyright case.
• A January 2023 internal memo by Microsoft's Brent Hecht called AI scraping 'an astonishing theft of unprecedented proportions' and 'the largest theft of labor in human history.'
• OpenAI's Nick Turley described chatbots as 'largely substitutive' and an existential threat to publishers; OpenAI internally called itself an 'existential threat' to news publishers.
• An internal Microsoft document warned its AI content strategy started a 'doom loop' threatening the web's content supply chain.
• Traffic-impact figures differ by source: TechCrunch says the Copilot answer engine cut NYT click-through rates by up to 93% versus Bing; Ars Technica cites Microsoft-recorded 83-93% click-through-rate drops for some news plaintiffs; 404…
• OpenAI datasets reportedly contained 91,692 copies of the plaintiffs' works and over 2 million nytimes.com documents; Project Mango held 160,903 unique works.
• Internal messages show a crawler 'hack' letting OpenAI bypass the NYT paywall, which Greg Brockman applauded ('ah, nice').
• Employees allegedly circumvented NYT paywalls and stripped copyright notices from training data.