OpenAI and Microsoft Executives Admit LLMs Are Built on Theft and Creating a Doom Loop
Bot Mutiny |
Unsealed court documents in the New York Times lawsuit reveal internal admissions from Microsoft and OpenAI that their AI models destroy the economic foundations of their own data suppliers.
The people building generative AI know exactly what they're doing. They're just hoping the courts don't care. Internal documents and depositions from Microsoft and OpenAI, unsealed on September 17, 2026, show executives describing their own technology as predatory. These admissions were first reported by 404 Media, based on an unredacted filing in the ongoing copyright lawsuit brought by the New York Times. The unsealed court filing contains statements that OpenAI and Microsoft previously fought to keep secret. In these documents, Microsoft executives don't use the sanitized language of "innovation" or "transformation." They call the process of training large language models (LLMs) an astonishing theft of unprecedented proportions. ## The Doom Loop
One internal Microsoft document admits the company has created a "doom loop." The filing states that AI products hoover up human-generated content and then cannibalize the clicks those creators rely on. This effectively kills the businesses that provide the data the AI needs to function. The document notes it's highly unusual for an end-product to threaten the economic foundations of its own essential suppliers. Yet, that's the reality for the LLM business. Microsoft executive Brent Hecht wrote that LLMs steal content without distributing economic value back down the supply chain. He warned this necessarily threatens the economic stability of creators. ## Cratered Traffic
The impact isn't theoretical. Microsoft CEO Satya Nadella testified under oath that after AI began ripping content from news sites, clicks to those original sources fell off a cliff. On Bing, traffic to these sites cratered by more than 90 percent. OpenAI's own software engineers testified that it doesn't matter how prominently they show links to sources. Users simply won't click them. OpenAI executives even described their company as an existential threat to news publishers. ## Intentional Bypassing
The filing shows that the theft wasn't just a side effect of broad scraping. It was a targeted effort. Documents reveal OpenAI created a specific hack to get around the New York Times paywall. When Greg Brockman, OpenAI cofounder, was told about it, his response was "ah, nice."
OpenAI policy director Jack Clark wrote that the company was creating systems that substitute for the labor of the people who define the culture of society. These companies aren't just building tools. They're building replacements for the people they're currently robbing. ## The Fair Use Defense
In public, OpenAI and Microsoft lawyers argue that training models on the entire internet is fair use. They claim they're building something fundamentally different from the human labor used to train it. The internal documents tell a different story. The executives admit they're using content in a way that almost no one intended, without compensation. They know they've built a product that destroys its own supply chain. They know it's a theft of labor. They just didn't want you to see them say it under oath.