Artificial intelligence has consumed vast stretches of the internet. Now, it’s after our books.

We're not talking about just facts or fragments of information, but language that has been carefully considered, edited, and arranged into something meaningful. Stories that authors carried inside themselves for years. Arguments built page by page. Our histories, painstakingly recorded so they wouldn't be forgotten. Books offer AI something the internet cannot: the movement of the human mind from one thought to the next. They teach a machine more than what we know. They teach it how we reason, explain, imagine, and write.

For decades, the internet appeared to offer an almost endless supply of knowledge. But lately, it's become increasingly crowded with recycled and unreliable information. In the race to build more intelligent machines, technology companies have begun looking back to one of humanity’s oldest tools: the book.

Anthropic, the company behind Claude, identified books as one of the highest-quality sources of language available for training its models. In 2024, the company launched Project Panama - an initiative internally described as an effort to “destructively scan all the books in the world.”

Yes, the process is as disturbing as it sounds.

Millions of books were reportedly acquired through wholesalers, used-book sellers, and other suppliers. But they weren’t bought to be read. Instead, they essentially entered an assembly line. Their bindings were removed. Their spines were cut away. Their pages were fed through industrial scanners. Once the machine had taken what it wanted, our words, whatever remained of the book was destroyed.

Some of the world’s most advanced machines are being built by cutting apart one of mankind's oldest forms of memory.

Why purchase and destroy physical books when digital editions already exist, you ask? In part, because digital books come with licenses, agreements, and restrictions. In other words, it gets messy. AI companies have already faced lawsuits over the use of pirated book collections in their training. Buying a physical copy offers a cleaner route: if the book is acquired legally, they can create an internal digital copy, and then argue that using its contents to train their AI model qualifies as fair use.

In Anthropic’s case, a federal judge found its use of books for AI training transformative, while treating the company’s earlier acquisition of pirated books as a separate infringement issue.

But what Project Panama created was not a library. At least not one that any of us could enter. These books weren’t scanned so readers could discover them or so libraries could preserve them. Their contents were absorbed into privately owned technology. The public can't browse the collection. Historians can't examine it. Authors may not even know whether their work is inside it.

The beauty of a physical book is it can be read, resold, borrowed, donated, inherited, or found decades after everyone has forgotten it existed. Once transformed into private training data, its words might remain - but our relationship to it fundamentally changes.

The question, then, isn’t just why these books are being destroyed. It’s what happens to the knowledge they contain once the books are gone. What happens when the machine is the only one left with a copy.

When the Original Disappears

Digitization, on its own, is not the enemy. Digital archives can protect fragile materials, expand public access, and preserve books long after their pages deteriorate. But preservation depends on more than just making a copy. It depends on who controls it, who can access it, and whether the original still exists so that we can verify it.

This is the power of a physical book. It lives in more than one place. Once released into the world, copies scatter across libraries, bookstores, classrooms, private collections, and homes.

This allows them to be compared and shared, or rediscovered. Knowledge spreads and becomes difficult to contain. No single company can alter every copy, close every library, or erase every edition at once.

A digital-only record is something else entirely. When knowledge exists solely on privately owned servers, public access can easily disappear behind a paywall, a forgotten password, or the collapse of a company. If the original has already been destroyed, how can anyone confirm that its replacement is complete or even accurate? Which edition was scanned? Which pages are missing? What images, notes, or context were lost along the way?

The danger isn’t necessarily that history will be rewritten all at once. The danger is in the fact that we might not even notice it happening. A missing page. An incomplete scan. A word copied incorrectly. Each loss may seem insignificant on its own, but together they create a version of the past that is shaped as much by what’s missing as by what remains. Eventually, the version we’re shown is the version we believe, no matter how incorrect.

Books have carried human memory through wars, censorship, political upheaval, and the rise and fall of empires. They have survived because they exist in many places and belong to many people. When one copy disappears, another remains.

But when the copy becomes the only version left, and that version belongs to someone else, how can we ever know what was lost, altered, or left out?

Who Gets to Write History?

The issue, ultimately, comes down to control.

History is shaped by those with the power to record the facts, to preserve them, and to decide which accounts survive. Governments have rewritten textbooks. Institutions have excluded uncomfortable truths. Entire communities have been reduced to footnotes, or erased altogether, because those in power controlled the narrative.

The same danger exists here. When our collective knowledge is concentrated inside systems owned by a small number of private companies, those companies gain unprecedented influence over what can be accessed and how it's presented. They don’t need to rewrite a book. All they have to do is decide which information appears in the search. AI makes that influence especially difficult to see. We’re given the answers without being shown where it came from.

This does not mean AI will inevitably be used to falsify history. But it does mean we’re building a system where falsification, omission, and distortion could become harder to detect and challenge. If the source material disappears, or stays locked inside private servers, we lose the ability to compare AI’s interpretation with what was actually written.

Our Final Thoughts

When we imagine a war on knowledge, we think of fire. Libraries burning. Books piled high in public squares. Something violent, loud, and impossible to ignore. But maybe this is how a knowledge war actually begins. Not with one dramatic act or public declaration, but with a gradual disappearance. Books vanish from the hands of the public. Ideas are moved somewhere we can no longer reach them. Ownership changes hands. Access slips away. And by the time we notice, the library still exists - it just no longer belongs to us.

The knowledge war isn’t a battle between books and technology. It is a battle for control over knowledge itself. Who decides what survives, what disappears, and which version of the truth the rest of us are allowed to see.

This is why the right to read has become our cause. Reading has never been just a pastime, and believe it or not, reading isn't passive. Books let us step outside the version of the world that's been handed to us. They allow us to experience ideas beyond our own and introduce us to lives we've never lived. They help us challenge what we’ve been taught and they give a voice to the voiceless.

We’re fighting for the freedom to follow our own curiosity - to choose which stories we read, which ideas we explore, and which questions we ask. We’re fighting for access to knowledge that can’t be controlled by a corporation, a government, or an algorithm deciding what is appropriate, relevant, or true. And we’re fighting to keep original works in the hands of the public, so we can always return to the page and read the words as they were written - not as a machine remembers them.

There's a reason why those who want to control people so often begin by controlling what we read. Books carry ideas across borders, survive generations, and outlive the people who try to ban them, burn them, rewrite them, or bury them. Protecting books means protecting our ability to question authority, confront history, form conclusions of our own. To think for ourselves.

Because when someone controls what we read, they do more than limit what we know. They can define what can be known all together. When they control the books, documents, and histories we depend on, their version becomes harder to question, and nearly impossible to disprove. If the only surviving copy belongs to them, so does the power to tell us what it said. 

The greatest weapon in a knowledge war isn't destruction. It’s control. And maybe that’s how this war is won: not by destroying every idea, but by owning the only version left. Control the source, and you control the story.