Newly unsealed court documents from the New York Times copyright lawsuit reveal that OpenAI and Microsoft internal teams explicitly warned their AI content strategies would trigger a self-defeating “doom loop” for web publishers, destroy their own data supply chains, and function as the largest theft of labor in history.
Nearly three years into a high-stakes legal battle between the New York Times, OpenAI, and Microsoft, newly unsealed court filings have pulled back the curtain on internal panic at the tech giants. The unsealed documents show that executives and engineers recognized early on that scraping vast swaths of copyrighted text without authorization or compensation threatened the very ecosystem that feeds large language models.
Internal Warnings of a Content Supply Chain Collapse
The court papers contain stark admissions from within Microsoft and OpenAI about the unsustainable nature of wholesale web harvesting. Among the most striking internal assessments came from Microsoft’s Director of Applied Science, Brent Hecht, who characterized the automated harvesting of data for ChatGPT and Copilot as the largest theft of labor in human history
and argued that corporate defenses made a complete mockery of the idea of ‘fair use.
Yet the filings demonstrate that the internal anxiety spanned across companies. OpenAI Policy Director Jack Clark noted that the industry was creating systems that substitute for the labor of the people that define the ‘culture’ of society.
Furthermore, internal OpenAI communications described ChatGPT as the modern newsstand,
where users obtain information without needing to visit the original publisher. OpenAI’s Nick Turley conceded that once a user receives a chatbot answer, there is no good reason to click
through to the source material.
Plummeting Referral Traffic and Regurgitation Risks
The economic impact on publishers is starkly quantified in the legal submissions. The New York Times highlighted Bing click-through rates on its articles, which cratered by 83-93% following the widespread rollout of AI-powered news summaries. OpenAI’s own media and economic experts independently estimated that search referrals across affected news sites may have fallen by as much as 60 percent due to features like Google’s AI Overviews and chat summaries.
Technical hurdles compounded these economic concerns. Employees acknowledged that GPT-4 memorized a ton of data and therefore will be insanely good at regurgitation,
posing severe copyright liabilities despite internal goals to prevent memorization. The filing documents instances where ChatGPT outputted long strings of text straight from the New York Times, Mercury News, The Denver Post, LifeHacker, and Eurogamer in response to user queries.
Despite Satya Nadella later acknowledging that anything that is paywalled should be licensed,
an OpenAI representative admitted in the unsealed record to being unaware of any systematic effort to detect or filter out paywalled content during model training.
The Ongoing Legal War and Political Shifts
The unsealed summary judgment requests highlight the central legal disagreement: whether harvesting decades of journalism constitutes protected fair use or a commercial substitution designed to usurp publishers’ revenues. While the New York Times argues that the defendants are deliberately replacing underlying journalistic labor, the litigation takes place against a shifting political backdrop.
Complicating the publishers’ position, the U.S. Department of Justice under the Trump administration recently intervened in the legal proceedings on the side of the technology companies. Government filings argued that The United States has a strong interest in this court rejecting any argument that training LLMs on copyrighted texts violates copyright law,
providing a major regulatory tailwind to OpenAI and Microsoft as the multi-year copyright battle proceeds.