atisusantiThe explosive growth of artificial intelligence, particularly generative AI tools that can produce human-like text and images, has sparked a curious phenomenon in the literary world: a significant surge in secondhand book sales. While this boom might seem like a resurgence of traditional reading habits, a closer examination, tracing back to a pivotal US court ruling, suggests a more complex and potentially unsettling driver: the insatiable appetite of AI for vast quantities of data, including printed texts. This connection, though not definitively proven for all booming sales, has sent ripples of unease through the bookselling community, raising questions about the future of literature and the ethics of digital data acquisition.
The idea that AI is fueling the secondhand book market gained traction following a 2025 US court ruling. In a landmark decision, Judge William Alsup ruled that an AI firm’s use of copyrighted books for training its AI software did not constitute a violation of US copyright law. This pivotal judgment stemmed from a lawsuit filed by three authors against AI firm Anthropic, known for its chatbot Claude. Judge Alsup’s reasoning centered on the concept of "exceedingly transformative" use, arguing that Anthropic’s repurposing of the authors’ works for AI training fell within legal parameters.
However, the unsealing of court documents last month revealed a more alarming aspect of this AI training process: books were not just being read, but in some cases, actively destroyed in the pursuit of digitizing and processing their content. Anthropic, in a statement, acknowledged its AI is trained on a diverse range of data, including publicly available web content, commercially acquired datasets, and internally generated data. While they insisted that their data acquisition programs do not target rare or antiquarian books for destruction, the revelation that books were being pulped to facilitate scanning has ignited considerable concern.
The internal company communications within Anthropic, as revealed by the unsealed documents, referred to the project of ingesting old books as "Project Panama." The ambitious and frankly chilling aim of this initiative was to "destructively scan all the books in the world." Destructive scanning is a process designed for industrial-scale digitization, where books are sent to specialized facilities. To achieve rapid scanning, the spine of a book is removed, allowing all pages to be processed quickly. The remaining shredded pages are then typically recycled. This method, while efficient for digitizing, fundamentally annihilates the physical book.
The term "Project Panama" and its underlying methodology are not entirely new to some within the bookselling community. Mark Manley of Barter Books, a well-known establishment, stated that while the name "Project Panama" is new to him, the reality of such projects has been a subject of much discussion on bookseller forums. He expressed that it’s difficult to pinpoint the exact destination of his sales, which have been remarkably diverse, ranging from obscure Latin texts to classic cowboy novels. This seemingly random pattern of sales, lacking any discernible rhyme or reason, further fuels speculation about AI-driven procurement.
Experts in the field of artificial intelligence suggest that this diverse subject matter is indeed a strong indicator of AI involvement. The argument is that unusual and rare texts, across a wide spectrum of topics, can provide novel and less common data points that are invaluable for improving the training of large language models (LLMs). LLMs are the foundational technology behind generative AI tools like chatbots, and their performance is directly correlated with the breadth and depth of the data they are trained on. The more varied and obscure the text, the more nuanced and capable the AI can become.
The legal landscape surrounding AI training and copyright differs significantly between the US and the UK. Professor Emily Hudson, an intellectual property specialist at Oxford University, highlighted this distinction. "The starting point in the UK is that all these acts of copying – creating the training library and doing the training – require the permission of the copyright owner," she explained. This means that the "transformative use" defense, which was central to the US ruling, might not offer the same protection under UK copyright law. The implications for AI firms operating in the UK are therefore more stringent, potentially requiring them to obtain explicit licenses for copyrighted material used in training.
For booksellers, this situation presents a profound ethical and practical dilemma. On one hand, they are witnessing an unprecedented surge in sales, which is undoubtedly beneficial for their businesses, especially in an era where many independent bookstores struggle to survive. On the other hand, the discomfort with the idea of books, some of which may hold significant historical or cultural value, being systematically destroyed for digitization is palpable.
Derek Walker, owner of Edinburgh bookshop McNaughtan’s, articulated this concern. "A recent academic text published in only 100 copies, 75 of which are already in libraries, may be very rare on the market – but it is perhaps not such a great loss if one copy is destroyed," he conceded. However, he quickly followed with a stark contrast: "But we have, and have sold, books which are for example the only known surviving example of an edition from the 18th century. It would be a much more significant problem if one like that were to be bought for destruction, having survived this long." The potential loss of unique historical artifacts and literary treasures weighs heavily on the minds of these custodians of the written word.
Despite these ethical reservations, there’s also a pragmatic acknowledgement that not every book needs to be preserved indefinitely in its physical form. Manley, from Barter Books, pointed out that recycling books is a practical solution for many titles that have become obsolete or are no longer in public demand. He also emphasized the immediate economic benefit: "Some may have ethical concerns about where the books end up and if they’re destroyed. But the world no longer needs five million copies of The Da Vinci Code. I’ve had books advertised for 20 years on the web which haven’t sold until now." This sentiment highlights the complex trade-off between preserving physical artifacts and the economic realities of the book trade, coupled with the potential for AI to unlock new value from what might otherwise be discarded or remain unsold.
The booming secondhand book market, therefore, appears to be a double-edged sword. While it offers a lifeline to struggling booksellers and a means for less sought-after titles to find new homes, it also raises serious questions about the long-term implications of AI’s data acquisition strategies. The notion of "Project Panama" and its destructive scanning methods serves as a stark reminder of the potential costs associated with the relentless pursuit of data for artificial intelligence. As AI continues its exponential development, the literary world, and indeed many other creative industries, will grapple with the ethical and practical challenges of balancing technological advancement with the preservation of cultural heritage. The future of books, both physical and digital, hangs in the balance, contingent on how societies choose to navigate this intricate intersection of AI and intellectual property.
Roy Suryo
George Anjay
Nakalama John
Ramos Leo
