AI Ethics

To Build an AI, They Are Burning the Books

The practice of destroying physical books for AI training data forces a reckoning with what cultural artifacts we are willing to sacrifice for innovation.

Imagine the sound of a book’s spine cracking for the last time. This isn't a scene from a dystopian novel; it’s a business practice. The act of destroying physical books after scanning them to train AI models is a stark illustration of the tension between technological advancement and cultural preservation, forcing us to ask what we are willing to erase in the name of progress.

Organizations, including the Internet Archive, have engaged in a process of mass digitization that sometimes ends with the physical destruction of the books. After a book’s text is scanned and converted into data for training large language models, the physical object itself can be seen as a liability—a thing that costs money to store and transport. From a purely logistical standpoint, shredding the paper is the final step in an efficient data-processing pipeline.

This practice has now drawn legal scrutiny. In a complaint to the Federal Trade Commission, a coalition of civil society groups argues that this destruction is not merely a tragedy for archivists but a potential violation of antitrust law. The argument is that by destroying the physical source material, a company prevents its competitors from ever accessing and scanning those same unique editions, effectively hoarding a finite resource and stifling competition in the AI development space.

What Is Lost When a Book Becomes Only Data?

A book is more than the text it contains. It is a physical artifact that carries its own history—the faded cover of a library copy, the weight of the paper, the pencil-thin margin notes left by a reader fifty years ago. When a book is destroyed after being scanned, its text may live on as a disembodied string of data, but its soul as an object is extinguished. This digital ghost preserves the information but erases the tangible connection to its past, much like keeping a photograph of an heirloom after casting the object itself into the furnace.

The destruction of books for data is a powerful symbol for a much larger cultural shift. We are increasingly asked to accept digital facsimiles as complete replacements for physical reality, from virtual meetings that replace shared space to digital media that replaces records, films, and photographs. Each instance raises the same question: does the convenience of the digital copy justify the erasure of the original? And who gets to decide which artifacts are worth preserving in physical form?

Ultimately, the debate over book destruction is a debate about the values embedded in our technology. It asks whether the future must be built upon the ashes of the past, or if we can find a way for innovation and preservation to coexist. If our most advanced tools for knowledge are trained on the remnants of culture they themselves helped erase, what kind of knowledge will they truly possess?

FAQ

Why are some organizations destroying physical books after scanning them? After a book is digitized to create training data for AI, the physical copy incurs storage and management costs. For large-scale scanning operations, destroying the book is often considered more economically efficient than storing, shipping, or reselling it.

What is the core argument of the complaint filed with the FTC? The complaint frames the destruction of books as an anticompetitive act. It alleges that by destroying the physical source, a company prevents competitors from using those same unique books, thereby consolidating a finite resource and harming competition in the AI market.

Isn't digitizing books good for accessibility? Digitization itself is widely seen as beneficial for making knowledge more accessible. The controversy is specifically about the subsequent destruction of the physical originals, which removes them as cultural artifacts and prevents anyone else from accessing that specific copy for any purpose, including competing AI training.