Court filings unsealed internal Microsoft and OpenAI documents. A Microsoft exec called AI scraping the largest theft of labor in human history. Another document called it a mockery of fair use. OpenAI staff warned of an existential threat to news publishers. Microsoft described a doom loop harming both models and the web. Click-through rates for news sites dropped by 51–94 percent. Chatbots often reproduced articles verbatim when prompted. Microsoft and OpenAI allegedly bypassed paywalls to scrape content. OpenAI used a dataset bound by non-commercial terms for training. News groups argue this proves substitution and undermines fair use claims. Microsoft disclaimed the exec’s comments as one employee’s view. News publishers say the evidence shows knowing wrongdoing.
What changed
Internal warnings about AI scraping were kept confidential until court filings unsealed them.
What it unlocks
Public visibility into Microsoft and OpenAI’s internal assessments of AI’s impact on news publishers.
- 83–93% drop in click-through rates for some news plaintiffs
- 51–94% drop for others
- 1.8 million NYT articles in dataset
Sources