Microsoft executive calls AI scraping “unprecedented theft”: Key evidence emerges in NYT copyright case

In the New York Times (NYT) copyright lawsuit against Microsoft and OpenAI, newly unsealed court documents revealed a sharp internal comment from a Microsoft executive: Microsoft’s Director of Applied Sciences Brent Hecht stated that using large models to “scrape” content from across the web for AI training constitutes“unprecedented theft”This criticism from within the industry has subtly shifted public opinion in the copyright litigation.

The exposed key statement

According to court documents unsealed on Thursday, The New York Times’ lawyers cited a statement by Hecht: “Hundreds of millions of people worldwide will soon view large models’ ‘sucking dry’ of all their works as ‘an unprecedented theft’—‘almost no one wants their creative content used this way, and no one is compensated for it.’” Plaintiffs regard this internal remark as strong evidence undermining the legality of Microsoft’s and OpenAI’s model training practices.

How Microsoft responded

The Orlando Sentinel, also a plaintiff in this case, reported that Microsoft portrayed Hecht in a filing as “someone hired to play contrarian” (someone on the payroll to play a contrarian), attempting to diminish the weight of this internal statement. Yet regardless, an active Microsoft executive publicly questioning the copyright foundation of his own company’s technology dealt a significant blow to Microsoft’s and OpenAI’s defense.

Case background: A lawsuit defining industry boundaries

The New York Times sued OpenAI and Microsoft in late 2023, accusing them of unauthorized use of its articles to train ChatGPTmodels, constituting copyright infringement. This litigation has drawn intense attention because its ruling could set a precedent for the entire generative AI industry: if the court determines that scraping content from the public internet without authorization constitutes infringement, nearly all large-model companies relying on such data for training would face substantial compliance costs and legal risks.

What this means for the AI industry

As more internal documents emerge through litigation, AI companies’ “use-first-ask-later” training approach is facing mounting pressure. For practitioners and users alike, the trajectory of this case warrants ongoing attention—it may directly affect future AI products’ training data sources, licensing costs, and even the viability of certain free features.

Summarize

An internal executive’s single word—“theft”—adds industry-insider corroboration to the AI copyright debate. When even Microsoft’s own personnel question scrape-based training, AI companies’ reliance on “fair use” as a defense is likely to grow increasingly untenable.

AI Dash — Discover the best AI tools