Court Docs: OpenAI Knew Book Piracy Was Illegal

Overview

Unsealed legal filings in the high-profile copyright lawsuit Authors Guild v. OpenAI have revealed that top executives at OpenAI and Microsoft were fully aware that their use of pirated books for AI training was illegal. The documents, released on September 21, 2026, paint a picture of a tech giant willing to bypass ethical and legal boundaries to win the artificial intelligence arms race.

The lawsuit, brought by The Authors Guild and numerous prominent writers, alleges that the companies illegally trained their products on copyrighted literary works. New briefs suggest that this was not an oversight but a deliberate strategy, with internal communications showing that leadership dismissed concerns about the impact on human authors as "acceptable economic disruption."

Internal Awareness of Piracy

The most significant revelation from the court documents is that OpenAI’s leadership knew they were using stolen content from Library Genesis (LibGen), a notorious shadow library site. According to the filings, this practice was disclosed as early as April 2019. During a presentation of an early version of GPT-3, CEO Sam Altman and Research Director Dario Amodei explicitly told Bill Gates and Microsoft CTO Kevin Scott that they were sourcing data from LibGen.

Internal memos indicate that OpenAI employees viewed the source with caution, not because they believed it was legal, but due to reputational risk. When Amodei acknowledged that using LibGen as a training set was "a bit sketchier," researcher Sam McCandlish clarified that his primary concern was optics, not legality. He noted that the headline "OpenAI uses copyrighted data from a sketchy Russian website" would be unfortunate if it appeared on platforms like Hacker News.

Despite these reservations, the company proceeded with the data scraping. The documents show that OpenAI was acutely aware that the books were being taken without permission, yet they prioritized the development of their models over the rights of the creators.

Disdain for Authorship and Labor

The filings also expose a chilling disregard for the professional livelihoods of writers. Jack Clark, OpenAI’s Policy Director, stated in May 2020 that the company’s work on AI creativity was designed to substitute for human labor. He explicitly warned that improvements in GPT models would cause genre fiction authors significant anxiety, anticipating that the technology would eventually replace their work on platforms like Amazon.

Clark further admitted that the company anticipated backlash from artists but planned to ignore those concerns and release their products regardless. He stated plainly, "Our work in this area will make people unemployed... There will be a point where a bunch of artists express worry about what we’re doing here and we’ll likely ignore their concerns and release anyway."

This sentiment was echoed by other team members. Tarun Gogineni, hired in 2022 to improve writing quality, openly discussed the goal of having GPT models autocomplete unfinished literary series, such as George R.R. Martin’s A Song of Ice and Fire. Gogineni expressed relief that even if an author died prematurely, the AI could finish their work, viewing this capability as a success metric rather than an ethical violation.

Evidence Destruction and Legal Strategy

Perhaps most damning is the evidence that OpenAI attempted to destroy its own records when faced with potential legal scrutiny. In the summer of 2022, a campaign internally dubbed "Project Clear" was initiated to remove traces of the pirated data.

On June 15, 2022, after noticing that mentions of LibGen were appearing across Google Docs, Slack, and GitHub, an employee asked how concerned the company should be about these references. That same evening, VP of Research Bob McGrew responded, stating, "Given how much OpenAI is in the news, now is the right time to excise Libgen from our systems and storage."

This coordinated effort to delete files suggests that the company recognized the illegality of its actions and sought to cover its tracks before the information became public. The plaintiffs argue that this demonstrates guilty knowledge, proving that the infringement was intentional and not incidental.

Impact on Authors and Culture

The lawsuit is brought on behalf of more than 18,000 members of The Authors Guild, including best-selling authors such as David Baldacci, John Grisham, George R.R. Martin, and Jodi Picoult. They argue that OpenAI’s actions pose an existential threat to the literary profession.

Mary Rasenberger, CEO of the Authors Guild, stated that the filings reveal a shocking disdain for writers. She emphasized that the company made repeated, intentional decisions to steal books rather than pay for them, fully knowing that their products would destroy the careers of authors. Rasenberger added that OpenAI pursued this mass piracy scheme despite clear evidence that it would degrade American culture by substituting human works with AI-generated content.

Gogineni’s own writings referenced in the briefs warned of "the death of the reader," predicting a future where machines create content for other machines. However, the plaintiffs view this not as inevitable progress but as a preventable harm caused by corporate greed and legal evasion.

Current Legal Proceedings

The case, formally known as Alter v. OpenAI and Microsoft, is part of multidistrict litigation pending in Manhattan. Lead counsel Justin A. Nelson from Susman Godfrey represents the class plaintiffs. The newly filed documents mark a turning point in the case, shifting the narrative from whether the data was used to how knowingly and intentionally it was obtained.

The plaintiffs are seeking partial summary judgment based on these undisputed facts. With more briefing expected over the coming months and a hearing scheduled for early 2027, the outcome could set a precedent for how AI developers handle copyrighted material. For the authors involved, these documents serve as validation that their concerns about intellectual property theft and economic displacement were well-founded.