Unsealed Court Docs Reveal AI Execs Feared 'Largest Theft of Labor' in History

Key points

• Microsoft Director of Applied Science Brent Hecht called the ingestion of copyrighted works an 'astonishing theft of unprecedented proportions' in a 2023 internal document.
• OpenAI President Greg Brockman reportedly responded 'Ah nice' when an engineer described a hack to bypass The New York Times paywall.
• Microsoft CEO Satya Nadella testified that paywalled content should be licensed, stating he would have required OpenAI to retrain models if he knew they used restricted content.
• Publishers argue AI chatbots are 'substitutive,' with internal data showing click-through rates to The New York Times falling 87% to 93% when users used Bing Chat instead of traditional search.
• The case centers on the fair use doctrine, with publishers arguing AI companies cannot claim no licensing market exists when they have signed deals with other publishers.

Background

This development follows a series of copyright lawsuits filed by major media outlets, including The New York Times, Ziff Davis, and numerous local newspapers, against OpenAI and Microsoft. Earlier in September 2026, the U.S. Department of Justice filed a statement of interest supporting OpenAI, arguing that AI training on copyrighted material is a transformative use essential for national security and U.S. AI leadership. The current unsealed documents provide the first direct evidence of internal skepticism within the tech giants regarding the legality of their training practices.

How outlets are covering it

The Hollywood Reporter and The Washington Post emphasize the stark contrast between the AI companies' public fair use defense and their internal admissions of theft. The New York Times highlights the 'doom loop' concern, where AI models might degrade in quality due to the lack of original human content. CNET notes that while Microsoft called the comments 'one employee's perspective,' the filing includes testimony from CEO Satya Nadella acknowledging that paywalled content should be licensed. The publishers argue that the AI products are 'substitutive' rather than transformative, directly undermining the market for journalism, a point the tech companies dispute by claiming their models offer new value through summarization and synthesis.

Why it matters

The outcome of this litigation could set a precedent for the entire AI industry, determining whether training large language models on copyrighted material without a license is legal. If the courts rule against the AI companies, it could force a massive shift in how tech firms acquire training data, potentially requiring widespread licensing agreements with media and publishing houses. Conversely, a ruling in favor of the AI companies could solidify the fair use doctrine for AI, allowing them to continue using vast amounts of copyrighted content without compensation, fundamentally altering the economics of journalism and content creation.

What to watch

The next phase of the litigation will focus on the four-factor fair use test, particularly the effect of the AI use on the potential market for the copyrighted works. Courts will likely examine the extent to which AI chatbots replace traditional search and news consumption. The unsealed documents may lead to further discovery requests regarding the specific methods used to circumvent paywalls and the internal communications between Microsoft and OpenAI regarding licensing obligations.