NEW YORK, September 17 — Internal Microsoft communications made public in unredacted court filings describe the mass scraping of copyrighted works to train artificial intelligence as “the largest theft of labor in human history,” according to reporting by news outlets including The Verge. The remark came from a Microsoft executive, and it surfaced in documents tied to a consolidated copyright infringement lawsuit brought by authors and other creators against Microsoft, OpenAI and related entities in the U.S. District Court for the Southern District of New York.
The filings were unsealed after a court ruling allowed portions of previously sealed material to become public.
What the documents show
Here’s what it means in plain terms. The people building generative AI needed text — books, articles, and other human-created work — to teach their models how to write. Much of that material was gathered by scraping the open web.
The court records now show that at least one executive inside Microsoft understood the practice in stark terms, even as the company pressed ahead with the technology. That gap between the internal language and the public posture sits at the center of the case.
Authors and artists argue the scraping amounts to uncompensated use of their work. AI companies have generally maintained that training on publicly available data is fair use under U.S. law. Microsoft has invested billions of dollars in OpenAI, the maker of ChatGPT, which launched in November 2022 and quickly became one of the most widely used AI services in the world.
The partnership helped push generative AI into mainstream products and workplaces, and it is now part of the legal record being examined in court.
Why this case matters beyond the courtroom
Think of it like a library that copies every book on its shelves to teach a new employee how to write — without asking the authors first. That is roughly the dispute in front of the court. The unredacted filings offer a rare window into how Microsoft’s own people weighed the ethical and legal implications of data scraping while the company continued building AI into its products.
The consolidated case is one of several high-profile matters that could shape the legal boundaries of AI development. The outcome could have far-reaching consequences for how AI models are built and what rights creators hold over their work.
No trial date has been set, and the litigation remains ongoing. What to watch next: whether more internal documents become public, and how the court’s eventual ruling on fair use reshapes the ground rules for every company training AI on human-created work.
