Rare Books Dismembered and Destroyed to Feed Artificial Intelligence Training Systems

In what represents a striking and controversial development in the rapid expansion of modern artificial intelligence, vast quantities of physical books—including rare and irreplaceable volumes—are being systematically acquired, dismembered, and scanned to provide training data for advanced machine learning models.

The extent of this secretive practice was brought to light by investigative journalist Emanuel Maiberg, a co-founder of 404 Media, during an interview with host François Picard on France 24's current affairs program "Spotlight." Maiberg's findings detail a covert industrial process that turns traditional reading materials into consumable digital inputs for proprietary technological systems.

According to the investigation, the pipeline relies on an expansive, globalized, and deliberately opaque supply chain. Books are procured in bulk through anonymous online marketplaces and commercial vendors, obscuring the identity of the end buyers and hiding the ultimate destination of the literary works from the public and original sellers alike.

Once these books are purchased, they are transported directly to large-scale scanning facilities designed for maximum throughput. Inside these industrial centers, the traditional physical integrity of the book is disregarded. Workers or automated systems slice away the bindings and spines of the books, reducing carefully crafted volumes to stacks of loose sheets.

By detaching the bindings, operators are able to feed the individual pages into high-speed scanning devices at rapid velocity. This destructive method enables facilities to ingest vast amounts of printed text in a fraction of the time required by non-destructive, manual digitization methods, effectively operating as an industrial processing line for literature.

The sheer magnitude of the initiative is immense, involving thousands and potentially millions of printed works that are fed into what has been described as an AI woodchipper. In the process, historic and culturally valuable physical artifacts are permanently ruined simply to accelerate the extraction of the text printed upon their pages.

This rapid conversion highlights a profound transformation in how human knowledge is preserved, owned, and accessed. For centuries, physical books have existed as durable cultural objects capable of circulating freely among readers, moving through community and academic libraries, and being handed down across multiple generations.

Once a volume is physically destroyed and absorbed into a proprietary artificial intelligence system, however, that open cycle is severed. The original text no longer exists as a tangible, shareable item. Instead, its contents survive purely in a virtual state, encoded deep inside proprietary model architectures controlled entirely by private corporations.

This structural shift places the ownership and dissemination of information into corporate hands. The outputs of these machine learning models, as well as the rules governing who can query or interact with them, are dictated by private technology firms, replacing open physical access with corporate gatekeeping.

The investigation illuminates a striking paradox within the modern technology sector. In their relentless pursuit to ingest ever-greater volumes of human knowledge to enhance machine capabilities, developers are simultaneously enabling the physical eradication of the very source materials that hold and document that collective wisdom.

Addressing the motivation behind this widespread destruction, Maiberg noted that the process should not be interpreted as an intentional, diabolical campaign to rid human society of physical books. Rather, he emphasized that the driving force behind the practice is something far more mundane: complete institutional indifference.

In the high-stakes race to develop increasingly powerful artificial intelligence systems, printed books are treated strictly as raw material to be optimized and consumed. The irreversible destruction of unique and historic physical volumes is viewed merely as an acceptable operational expense necessary to speed up computational data collection.

Maiberg argued that this operational mindset reflects a broader and concerning concentration of informational power. A similar pattern previously occurred with the emergence of major search engines and dominant social media platforms, which gradually centralized control over how users discover and navigate information online.

Artificial intelligence, however, takes this centralization of power to an unprecedented level. Rather than simply directing users to existing independent sources or hosting public discussions, AI systems absorb the underlying source material entirely and present synthesized answers through their own private interfaces.

The revelations have prompted serious concern across the cultural and academic sectors, where educators, researchers, and archivists have long prioritized the preservation of physical heritage. The findings raise urgent questions about the future governance of human knowledge as private infrastructure increasingly mediates access to historical and literary records.

As artificial intelligence systems continue to absorb the world's accumulated wisdom while destroying the physical artifacts that preserved it, society faces a fundamental dilemma regarding who will ultimately control the technological infrastructure through which future generations can access human thought.