On September 18, 2026, unredacted court filings unsealed in the U.S. District Court for the Southern District of New York dealt a devastating blow to the legal defenses of Microsoft and OpenAI in their multi-billion-dollar copyright dispute with The New York Times.
The litigation, initially filed in late 2023, centers on whether foundation model developers have the legal right to scrape millions of copyrighted news articles, investigative investigations, and opinion columns without licensing fees or consent. In their public court pleadings, Microsoft and OpenAI have steadfastly argued that training neural networks on public text constitutes classic 'fair use'—analogous to a human reader learning grammar and style from a library book.
However, the newly unredacted internal discovery materials tell a starkly different story. In an internal 2024 email exchange between senior Microsoft engineers and corporate vice presidents regarding training datasets for Copilot and Bing, a senior Microsoft director wrote candidly: 'Let us be intellectually honest with ourselves: what the industry is doing with unconsented web scraping is, in economic reality, the largest theft of labor in human history. We are extracting the creative output of millions of journalists and programmers to build products designed to replace them.'
In separate unsealed Slack messages produced by OpenAI during discovery, an executive in the product research division expressed deep alarm over the launch of SearchGPT, stating: 'Publishers have every right to be furious. We are building an interface that summarizes their investigative scoops directly in our chat pane, eliminating their ad impressions and subscriptions. For their business models, this is an existential economic threat.'
Legal analysts emphasize that these admissions could prove fatal to the defendants' fair-use affirmative defenses. Under the fourth statutory factor of the fair use doctrine (17 U.S.C. § 107), courts must evaluate 'the effect of the use upon the potential market for or value of the copyrighted work.' When a defendant's own executives acknowledge in writing that their product will destroy the market for the original works, establishing fair use becomes nearly impossible.
The unredacted disclosures have caused immediate fallout across the publishing industry. Guilds representing thousands of digital writers, photojournalists, and academic authors have filed motions to consolidate their class-action claims with the New York Times docket, seeking statutory damages reaching $150,000 per willfully infringed work.
As federal judges prepare for oral arguments on summary judgment later this year, the unsealed records have fundamentally altered the balance of power, making an industry-wide licensing pact or legislative intervention by the US Congress almost inevitable.