An unredacted court filing unsealed this week in the New York Times versus OpenAI copyright lawsuit contains a series of admissions from Microsoft and OpenAI executives that amount to a rare, documented acknowledgment of what critics have argued for years: large language models were built on stolen content, are cannibalizing the websites that provided that content, and are destroying the economic foundations of the news industry.
What Was Unsealed and Why It Matters
The filing, submitted by lawyers for the New York Times as part of a summary judgment request in the years-long copyright case, draws on internal Microsoft documents, sworn testimony, and company policy papers that had previously remained sealed or heavily redacted at the request of both Microsoft and OpenAI. It was discovered by Jason Kint, CEO of Digital Content Next, a trade organization representing digital media companies who has been tracking major AI copyright litigation.
The material is damning not because it reveals something hidden from public scrutiny, but because it shows that the companies themselves made these admissions internally and under oath. The language is unusually direct for corporate paperwork. One internal Microsoft document described generative AI products as having created a "doom loop" that is killing "the entire web."
The Doom Loop: Steal, Cannibalize, Destroy
The internal Microsoft document lays out a mechanism that is straightforward and brutal. AI products first ingest content from creators and publishers without meaningful compensation. They then return summaries and answers that replace the need for users to click through to the original sites, draining the advertising revenue and traffic that those sites depend on. The result is a collapsing supply chain. The document itself stated: "Our AI content strategy has started a 'doom loop' that will hurt the performance of our models and the entire web at the same time. It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its content supply chain."
Microsoft CEO Satya Nadella testified under oath that after Microsoft incorporated New York Times and other news content into its AI systems, clicks to those news sites cratered, falling by more than 90 percent on Bing. The document acknowledged that LLMs steal content "without ways of distributing economic value down the supply chain, which necessarily threatens the economic stability of those who create the content."
The Paywall Hack and the Link Problem
Among the more pointed admissions in the filing is evidence that OpenAI developed a workaround to bypass the New York Times paywall. When shown the evidence, OpenAI cofounder Greg Brockman responded with a brief "ah, nice," according to the filing.
That incident sits alongside a broader pattern documented in the filing. An OpenAI software engineer testified that "no matter how prominently we show the links, users won't click." OpenAI's own policy director, Jack Clark, wrote that the company was "creating systems that substitute for the labor of the people that define the culture of society." A separate Microsoft policy document stated that generative AI could "significantly disrupt the employment of the very people who generated the data on which the foundation model was trained," and described LLMs as "a product that destroys its supply chain."
The Fair Use Defense Meets Internal Reality
OpenAI and Microsoft have been arguing in court that their model training constitutes fair use and is transformative under copyright law, presenting their products as fundamentally different from the human labor they were trained on. The unsealed filing undercuts that argument by showing that the companies' own executives understood the dynamics at play and described them in stark terms.
Microsoft's internal language is telling. The company framed the situation as "an astonishing theft of unprecedented proportions" and "the largest theft of labor in human history," adding that "almost no one intended for content they created to be used in this fashion, nor are they compensated for its use." OpenAI, meanwhile, described itself as an "existential threat" to news publishers.
What the Filing Changes
None of these admissions would surprise anyone who has followed the development of generative AI and the anxiety it has caused among publishers, writers, and artists. What the filing changes is the evidentiary record. These are not critic quotes or outside analysis. They are internal documents and sworn testimony from the executives building the systems, acknowledging the mechanisms of harm they created and the economic damage those mechanisms are causing.
The case continues through the court system. But the unsealed filing has shifted the public conversation from whether AI companies benefited from unlicensed training data to what, precisely, they knew about the consequences and when they knew it. For the news industry, that distinction matters enormously, both legally and commercially.