Newly unsealed court documents from the copyright lawsuit filed by The New York Times reveal that high-level executives at Microsoft and OpenAI privately described their own business models as “labor theft” and a potential “existential threat” to the news industry. The records, made public on September 17, 2026, provide a sharp contrast to the companies’ public legal defense, which maintains that training AI models on copyrighted data is “fair use” under current law.
In a series of internal communications from 2023, Microsoft Director of Applied Science Brent Hecht characterized AI data scraping as the “largest theft of labor in human history.” Hecht further warned colleagues that successfully arguing for a fair use exemption for these practices would “make a complete mockery of the idea of fair use.”
The ‘Existential Threat’ and Project Mango
The unsealed filings also highlight internal memos from OpenAI leadership. Nick Turley, the Head of ChatGPT at OpenAI, wrote in a 2023 memo that the company’s AI products are “largely substitutive, period” for the original content they ingest. Turley explicitly stated that the technology posed an “existential threat” to publishers by replacing the need for users to visit original source websites.
Central to the collaboration between the two companies was a joint effort known as “Project Mango.” While the companies have publicly argued that their tools act as a “copilot” to assist users, internal data suggests a different outcome for content creators. Microsoft’s own internal metrics showed that click-through rates (CTR) to news publishers plummeted by 83% to 93% when users received answers through Copilot instead of traditional search results.
Bypassing Paywalls
The litigation has also brought to light how the companies handled subscription barriers. Internal logs show that OpenAI President Greg Brockman was informed of a technical “hack” used to bypass The New York Times’ paywall during the model-training process. Brockman reportedly responded to the development with “Ah, nice.” Microsoft spokesperson Alex Haurek has since distanced the company from Hecht’s “theft of labor” comments, stating they reflect only “one employee’s individual perspective” and do not represent the corporation’s official stance.
The ‘Doom Loop’ for Content
The unsealed briefs outline a technical concern internal teams referred to as a “doom loop” for the web. Documents suggest that by destroying the economic foundation of journalism through traffic redirection, AI companies risk polluting the “supply chain” of human-generated content required to train future versions of their models. If the primary sources of high-quality data are driven out of business, the AI models would eventually be forced to train on AI-generated “synthetic” data, leading to model degradation.
The plaintiffs argue that these internal admissions of “substitutive” behavior could lead to significant statutory damages. Under copyright law, proving “willful” infringement—knowledge that one’s actions are infringing while continuing the practice—can drastically increase the financial penalties per work. As the case proceeds, these internal warnings that the business model would make a mockery of the idea of fair use are expected to become central to the court’s determination of whether the companies acted in good faith.
