Claude and a Million Dead Books: How AI Devoured Our Literary Heritage

Remember that quiet November of 2022, when the world didn't yet know what AI hallucinations were? OpenAI didn't just release a product back then. It fired a starting pistol, and the sound sent every Silicon Valley giant running, leaving their cozy metaverses behind. In this race, Anthropic's Claude became one of the leading contenders for the crown. But behind the polished facade of a polite, safe AI lies a graveyard of millions of books that nobody gave permission to use. Let's be honest: Claude exists in its current form only because the industry decided to ignore the rules of decency for the sake of speed.

Today's AI industry resembles an era of wild capitalism, with datasets standing in for gold mines. Once it became clear that ChatGPT wasn't just a toy but the foundation of a new economy, ethics fell far down the list of priorities. Text from Reddit or Wikipedia isn't enough to teach a model to reason. It needs complex structures, a rich vocabulary, and deep context, and those can only be found in quality literature. That's how millions of copyrighted books became a "training set" without their creators' consent. You didn't really think neural networks learn from public-domain children's fairy tales, did you?

Anthropic often positions itself as the "good guys" of the AI world, with a focus on safety and ethics. The irony is that even the "safest" models are built on a foundation of questionable content. Datasets like Books3, which contain hundreds of thousands of titles from shadow libraries, became the secret ingredient that let Claude catch up with Sam Altman's products and, in some respects, overtake them. For the corporations, the math was simple: either you use everything that isn't nailed down, or your competitor does it first and captures the market. By that logic, books are just coal for the furnace of progress.

Why does this matter right now? We're approaching the point where "human" data simply runs out. Neural networks have already read almost everything we've written over the past few centuries. Now authors, from novelists to technical writers, are discovering that years of their work have become free fuel for systems that may one day replace them. This isn't just content theft. It's a fundamental shift in how we understand intellectual property. You used to buy a book to read it; now corporations take it to teach a machine to imitate your style and the way you think.

Lawsuits from authors such as Sarah Silverman and George R.R. Martin are just the tip of the iceberg. The problem is that the legal system is sluggish, while the AI industry moves at the speed of light. While courts spend years deciding whether training a neural network counts as "fair use," the models are already trained, the weights are saved, and billions of dollars in investment have been spent. Anthropic and other players bet that winners don't get judged. Or, at the very least, that copyright infringement fines will be a drop in the ocean compared with their future market capitalization.

In the end, we're left with a strange symbiosis. Claude can analyze a complex legal document for you or write an essay in the style of Proust precisely because it "swallowed" thousands of similar texts without asking. We got an incredible tool, but the price of building it is the devaluation of human labor itself. Books didn't just serve as a foundation; they were ground into digital mincemeat and reshaped into new, easily consumable interfaces. And now we have to live with that, enjoying the fruits of this intellectual expropriation.

The bottom line: Anthropic and OpenAI built their empires on data that didn't belong to them, and there's no going back. Can the industry survive if it eventually has to pay the bill for every book it has "read"?