In a landmark legal battle that could reshape the future of AI and copyright law, OpenAI has emerged victorious in a critical discovery dispute against authors, potentially shielding AI companies from billions in damages. But here's where it gets controversial: the ruling hinges on whether OpenAI’s denial of willful copyright infringement waives its attorney-client privilege, a decision that has sparked fierce debate among legal experts. Let’s break it down.
The case revolves around OpenAI’s use of pirated books to train its GPT models. In 2018, an OpenAI employee downloaded copyrighted books from LibGen, a shadow library, and created two datasets—“books 1” and “books 2”—to train now-discontinued GPT models. OpenAI later deleted these datasets in 2022, initially citing “non-use” as the reason. However, authors and publishers, including Sarah Silverman, sued, alleging willful copyright infringement. The plaintiffs’ lawyers argued that OpenAI’s actions were deliberate, potentially exposing the company to damages of up to $150,000 per work, compared to just $200 for unintentional infringement.
The crux of the controversy lies in whether OpenAI’s denial of willful infringement opens the door to privileged internal communications. Magistrate Judge Ona Wang initially ruled that by denying these allegations, OpenAI waived attorney-client privilege, effectively forcing the company to disclose its motivations for deleting the datasets. This decision sent shockwaves through the legal community, as it could set a precedent for AI companies to lose privilege protections whenever they deny intentional copyright violations.
OpenAI fought back, enlisting Supreme Court veteran Lisa Blatt, who warned that upholding the ruling would “eviscerate” privilege assertions in copyright cases involving state of mind—a critical analysis for determining intent. On Friday, U.S. District Judge Sidney Stein reversed Wang’s decision, arguing that denying willful infringement is not the same as asserting a good faith defense. This reversal keeps OpenAI’s internal communications shielded, but it raises a provocative question: Are AI companies being held to a fair standard when it comes to copyright law, or are they exploiting loopholes to avoid accountability?
And this is the part most people miss: while OpenAI won this battle, the war is far from over. Authors’ lawyers are pivoting to a new strategy, arguing that the mere act of illegally downloading copyrighted works—regardless of whether they’re used—constitutes infringement. This approach builds on a recent ruling in a separate case against Anthropic, where U.S. District Judge William Alsup stated, “That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for the theft.” Anthropic ultimately settled for $1.5 billion, and this precedent is now fueling the OpenAI case.
Here’s the bigger picture: AI companies have raised billions by training models on pirated content, yet the legal framework remains murky. OpenAI’s use of “books 1” and “books 2” is just one example of a widespread practice that has yet to be fully addressed by the courts. As AI continues to evolve, so too must the laws governing its development. But at what cost to creators? And should AI companies be allowed to profit from potentially illegal practices?
What do you think? Is OpenAI’s victory a win for innovation, or does it undermine the rights of authors and publishers? Let’s debate in the comments—your perspective could shape the conversation.
