Anthropic Faces Billions in Damages Over Alleged Copyright Infringement

• Anthropic faces a class-action lawsuit for allegedly using pirated books to train its AI model, Claude.
• The lawsuit could result in billions of dollars in damages if Anthropic loses.
• Judge William Alsup ruled that training AI on lawfully acquired books is “fair use,” but the case against Anthropic remains unresolved due to disputed book origins.
• Potential damages could range from $1 billion to $1.05 trillion, depending on the number of works and willful infringement.
• The case may have implications for similar lawsuits, including the OpenAI dispute with authors and publishers.

Anthropic, California-based prominent AI lab, is facing a class-action lawsuit that poses a significant threat to its future. The lawsuit alleges that the company used potentially pirated books to train its Claude large language model, which could result in billions of dollars in damages if the company loses the case. This legal battle may have far-reaching consequences for Anthropic’s prospects.

According to court documents, Anthropic allegedly downloaded millions of copyrighted works from shadow libraries such as LibGen and PiLiMi. The purpose was to train its AI models and create a vast digital library of books, aiming to include “all the books in the world” and preserve them indefinitely.

The plaintiffs, authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, claim that millions of these works were obtained from piracy websites, directly infringing on copyright law. They allege that Anthropic’s actions violate the rights of authors and creators, and are seeking damages for the unauthorized use of their works.

Judge William Alsup recently ruled that training AI models on lawfully acquired books constitutes “fair use,” meaning AI companies don’t need a license from copyright holders for such training. This decision is seen as a significant victory for the AI sector, potentially setting a favorable precedent for AI development and usage.

While Judge Alsup ruled in favor of AI companies using lawfully acquired books for training models, the case against Anthropic remains unresolved due to the disputed origin of the copyrighted books. The judge indicated that a separate trial will focus on whether Anthropic used pirated materials and determine the resulting damages, potentially impacting the company’s liability.

Many AI companies have allegedly scraped content from piracy sites like LibGen, where books are uploaded without author permission or payment. “The problem is that a lot of these AI companies have scraped piracy sites like LibGen … where books have been uploaded in electronic form, usually PDF, without the permission of the authors, without payment,” says Luke McDonagh, an associate professor of law at LSE.

Luke McDonagh clarifies the judge’s stance on AI training data. “The judge seems to be suggesting that if you had bought a million books from Amazon in digital form, then you could do the training, and that would be legal, but it’s the downloading from the pirate website that is the problem, because there’s two things, there’s that acquiring of the copy, and then the use of the copy.”

According to Santa Clara law professor Ed Lee, the ruling could have severe consequences for Anthropic. “At least the potential for business-ending liability” is a possibility, Lee warns, due to the lawsuit.

Legal experts believe the plaintiffs are likely to pursue statutory damages, which can range from $750 to $150,000 per work, depending on whether the infringement is deemed willful. If the court finds that Anthropic knowingly violated copyright law, the fines could be substantial, potentially reaching billions of dollars, even at the lower end of the scale.

The potential damages in the lawsuit against Anthropic are uncertain, but they could be substantial. The number of works included in the class action and the jury’s finding on willful infringement will play a significant role in determining the damages.

According to Ed Lee, even a relatively low damages award could be in the range of $1 billion to $3 billion if 100,000 works are included in the class action. This figure is comparable to the largest copyright damage awards on record and could exceed Anthropic’s current annual revenue of $4 billion, potentially having a significant impact on the company’s financial situation.

According to estimates, Anthropic could face a liability of up to $1.05 trillion if found guilty of willfully pirating 6 million copyrighted books. Fortune’s request for comment from Anthropic did not receive an immediate response.

The company has previously stated that it “respectfully disagrees” with the court’s decision and is considering its options, including potentially appealing the ruling or negotiating a settlement. A trial is set for December 1, marking a significant legal milestone as the first certified class action case against an AI company regarding copyrighted materials.

The outcome of this case could have implications for similar lawsuits, including the ongoing dispute between OpenAI and numerous authors and publishers. Courts seem to be leaning towards allowing AI companies to make fair use arguments, but there’s a legal divide regarding the acquisition of copyrighted materials from shadow sites.

In a recent copyright case against Meta, Judge Vince Chhabria ruled that the transformative purpose of AI use can legitimize earlier unauthorized downloading. According to McDonagh, this ruling suggested that the positive, transformative use of works could “correct” the initial problematic acquisition.

In contrast, Judge Alsup viewed the downloading of books from unauthorized shadow libraries as “inherently wrong,” implying that even if AI training use might be considered fair use, the initial acquisition of works was illegitimate and warrants compensation. This differing perspective highlights the complexity and nuance of copyright law as it applies to AI development.

The judges in the Meta and Anthropic cases had differing opinions on whether AI-generated outputs could be seen as competing with the original copyrighted works used in their training data. Judge Chhabria noted that if competition was proven, it could undermine a fair use defense, but found that the plaintiffs in the Meta case didn’t provide enough evidence of market harm.

In contrast, Judge Alsup concluded that generative AI outputs don’t compete with the original works at all. This discrepancy highlights the ongoing debate and complexity surrounding AI companies and copyrighted materials.

The issue has also become increasingly politicized, with the current administration pushing for broad fair use protections to allow AI companies to use copyrighted materials for training. This effort aims to maintain US leadership in artificial intelligence. McDonagh believes the case against Anthropic is unlikely to bankrupt the company, speculating that the administration would intervene to prevent a ruling that could harm an AI company.

Judges tend to be cautious when issuing rulings that could lead to bankruptcy, considering the potential impact on the company and its stakeholders. Courts often weigh the legal basis and necessity of such actions before making a decision.

“The U.S. Supreme Court, at the moment, seems quite friendly to the Trump agenda, so it’s quite likely that in the end, this wouldn’t have been the kind of doomsday scenario of the copyright ruling bankrupting Anthropic,” McDonagh said. With Anthropic valued between $60 and $100 billion, paying a couple of billion dollars to authors would not likely bankrupt the company, suggesting a more manageable outcome.

ⓘ As part of our ongoing support for startups and SMEs, LAFFAZ Media publishes feature and resource articles that may include references and links to external websites. These inclusions are selected at our editorial discretion to provide valuable information to our readers. LAFFAZ Media does not control, endorse, or assume responsibility for the content or practices of external websites. For more details, please refer to our Terms and Conditions.