The Unseen Labor of AI: A Theft in Plain Sight

The unredacted filings in the ongoing copyright lawsuit against OpenAI and Microsoft reveal a stark reality: the use of copyrighted material to train generative AI models is not just an issue of fair use, but a brazen act of theft. Top executives at both companies have made admissions that paint a picture of a complex web of scraping, bypassing paywalls, and stripping copyright notices – all in pursuit of training data.

The scale of the copying is astonishing, with over 91,692 copies of works published by The New York Times alone contained within OpenAI’s mid-training datasets. This is not just an issue of copyright infringement; it’s a fundamental challenge to the notion of intellectual property in the digital age. As Brent Hecht, Microsoft’s director of Applied Science, noted in a January 2024 internal memo: “It is highly unusual that an end-product threatens the economic foundations of its essential suppliers.”

The implications are far-reaching and have significant consequences for the publishing industry as a whole. Generative AI models, touted as revolutionary tools for content creation, are in reality substitutes for human writers and journalists. Microsoft CEO Satya Nadella testified under oath earlier this year that conversing with chatbots “has substituted… giving you the information right there on the website on the AI platform versus needing to go to the underlying source.” This is not transformation; it’s substitution.

The fair-use test, relied upon by OpenAI and Microsoft to justify their actions, is no longer a viable defense. Judges have been largely favorable to AI companies’ arguments that training constitutes “fair use,” but top executives’ quotes reveal a disturbing disconnect between these claims and the reality of how generative AI models are being used.

The complex relationships between tech giants, publishers, and journalists are laid bare in the Microsoft document’s description of the “content supply chain.” The admission that companies like OpenAI and Microsoft will bypass paywalls undetected, stripping copyright notices from training data, raises serious questions about the ethics of AI development. Hecht described this as “an astonishing theft of unprecedented proportions… the largest theft of labor in human history.”

The Trump administration’s recent brief in defense of OpenAI’s unlicensed use of copyrighted material is a worrying sign that this issue will not be taken seriously by those in power. The lawsuit itself, now in its third year, has been met with resistance from AI companies who claim to be innovators but are in reality profiteers.

The future of publishing hangs in the balance: will we see a new era of “fair use” that permits companies like OpenAI and Microsoft to continue scraping copyrighted material without consequences? The filing details how these companies acquired plaintiffs’ content, including scraping it from the Bing Index. This is not just an issue of copyright infringement; it’s a fundamental challenge to the notion of intellectual property itself.

As we move forward, it’s essential to remember that the labor of journalists and writers is not just being stolen – it’s being replaced. Generative AI models are substitutes for human writers and journalists who have spent years honing their craft. The admissions from top executives at OpenAI and Microsoft reveal a stark reality: we are witnessing a seismic shift in the way we consume information, but also in the way we treat the people behind that information.

The unredacted filings in this lawsuit serve as a wake-up call for all of us who care about the future of publishing. It’s time to take a hard look at the role of AI in our industry and ask difficult questions: what does it mean to create content when machines can do it cheaper? What is the value of human labor in an age of automation? The answers will not be easy, but one thing is certain – the status quo is no longer tenable.