Photo: Can Ceylan / pexels.com

It sounds almost absurd: While the world is talking about the future of artificial intelligence, tech companies are looking deep into the past—into dusty archives, old libraries, and forgotten boxes of books. That is exactly where they are searching for something that is extremely valuable today: knowledge.

But this trend is increasingly sparking debate. Because what sounds like harmless digitization at first glance actually raises some pretty big questions.

Old Books, New Power: Why AI Suddenly Needs Antiquarian Bookstores

In the past, antiquarian books were something for collectors, historians, or lovers of yellowed pages. Today, they have taken on an entirely new role: they are a raw material.

AI systems need enormous amounts of text to learn how language works. The more diverse, the better. And this is exactly where old books come into play. They contain phrasing, styles, and content spanning centuries—a vast treasure trove of data that modern systems can use to become “smarter.”

The problem: Many of these books aren't readily available. Some are in libraries, while others are protected by copyright or, at the very least, their legal status regarding free use isn't clear. And that's exactly where the excitement begins.

Digitize, scan, train—but who actually gave their consent?

The technology itself is simple to explain: Books are scanned, texts are made digitally readable, and then they are fed into AI model training systems. For companies, this is a logical step—after all, every AI needs “food” to develop.

But the crucial question is: Are they even allowed to do that?

Much of this content was never intended to end up in machines that later generate text, images, or even entire books. Authors, publishers, and rights holders view this with skepticism. For while some speak of innovation, others speak of use without clear consent.

A tension is emerging between technological progress and the protection of creative work. And it is precisely this tension that is becoming increasingly relevant from both a legal and a societal perspective.

Between the Library and Silicon Valley: A Conflict of Systems

On one side are libraries and archives that seek to preserve knowledge and make it accessible. On the other side are tech companies that use this knowledge as the basis for new digital products.

Both sides argue in favor of a worthy goal: access to knowledge. Yet their approaches could hardly be more different.

Some say, “Knowledge belongs to everyone.”
Others say, “But not without rules.”

And this is precisely where a conflict arises—one that is far from being resolved. After all, AI is evolving faster than laws and social norms can keep up.

Why Old Books, of All Things, Have Suddenly Become Such a Hot Topic

The special appeal of antiquarian books lies in the fact that they often date from times when different styles of language were in use. This makes them valuable for AI systems because they help them understand language not only in a modern context, but also from a historical and multifaceted perspective.

At the same time, however, an uncomfortable question arises: If a book from a library is digitized and used for AI training—is that still purely preservation, or is it already commercial exploitation?

This uncertainty means that the field is currently undergoing significant change. Companies are experimenting, libraries are responding cautiously, and policymakers are trying to catch up.

Looking Ahead: Who Will Control the Knowledge of the Future?

The real crux of this development is not technical, but social. It is not just a matter of how AI learns, but also of who owns the knowledge it uses.

And it is precisely this question that will be crucial in the coming years. After all, the better AI becomes, the more its quality depends on the data it receives—and under what conditions.

Who Owns the Past in the Age of AI?

This trend feels a bit like a digital gold rush: Everyone wants quick access to as much “raw material” as possible, but hardly anyone likes to talk about where it comes from.

What’s particularly problematic here is the matter-of-fact way in which old books are suddenly treated as freely available training data. Just because something is old doesn’t automatically mean it’s in the public domain. This line of reasoning is convenient—but legally and morally rather shaky.

To put it provocatively: The AI industry is currently building the world’s most intelligent technology—and in doing so, it sometimes stumbles over the simplest questions of consent and respect.

And that is exactly why this issue isn't going away. On the contrary: it's getting bigger. And more uncomfortable.