# OpenAI and Microsoft's doom loop for the web

> Unsealed court documents reveal the companies knew their AI training would destroy publishers and damage their own products.
- Title: OpenAI and Microsoft's doom loop for the web · The Verge
- Summary: Unsealed court documents reveal the companies knew their AI training would destroy publishers and damage their own products. OpenAI and Microsoft's own…
- Keywords: openai, microsoft, new york times, lawsuit, ai training, copyright, court documents, chatgpt, copilot, fair use, content scraping, publishers
- Source: The Verge — https://theverge.com/ai-artificial-intelligence/997633/openai-microsoft-chatgpt-ai-new-york-times-doom-loop-theft-google-zero
- Author: Terrence O'Brien
- Published: 2026-09-18T21:07:24+00:00
- Read time: 1 min
- Topics: openai, microsoft, new york times, lawsuit, ai training, copyright, court documents, chatgpt, copilot, fair use, content scraping, publishers
## Court docs expose AI companies' knowledge
Unsealed documents from the New York Times lawsuit show OpenAI and Microsoft knew their AI systems posed an existential threat to publishers and would damage the web itself.
## Microsoft called it the largest theft ever
Microsoft's Director of Applied Science Brent Hecht described the scraping of data to train AI as the largest theft of labor in human history and a complete mockery of fair use.

> "the largest theft of labor in human history"
## They created a doom loop
Internal Microsoft documents admitted that AI chatbots have created a doom loop that hurts both the models and the web by replacing the need to visit original sources.

> "Our AI content strategy has started a 'doom loop'"
## Profit motive over ethics
OpenAI cofounder Greg Brockman wrote he was deeply motivated by the gazillions of dollars he hoped to gain by commercializing the technology, revealing profit as the driving force.

> "deeply motivated by the gazillions"
## ChatGPT memorizes and regurgitates content
OpenAI knew ChatGPT would reproduce copyrighted material verbatim, with employees admitting GPT-4 memorized data and would be insanely good at regurgitation.

> "insanely good at regurgitation"
## Paywalled content wasn't protected
Despite Satya Nadella saying paywalled content should be licensed, OpenAI admitted it made no effort to detect or remove paywalled articles from training data.
## AI replaces the need to visit sources
OpenAI's Nick Turley acknowledged that once users get an answer from ChatGPT, there is no good reason to click through to the original source, destroying referral traffic.

> "no good reason to click"
## AI destroys its own supply chain
Microsoft acknowledged that large language models destroy their own supply chain because they substitute for the training data they depend on, creating an unsustainable system.

> "destroys its own supply chain"
## Key takeaway

OpenAI and Microsoft's own internal documents show they knowingly created a self-destructive system that harms content creators and their own supply chain, but pursued it anyway for profit.