AI is out of data. Now it’s burning books

In 1994, Jeff Bezos started Amazon as a small online bookstore. The idea was simple. He used digital logistics to give more people access to books. Three decades later, his company runs a quiet warehouse in Las Vegas. There, industrial machines slice the bindings off rare books. Workers feed the loose pages into scanners and dump the shredded remains into recycling bins.An investigative report from 404 Media exposed this operation. An antiquarian book dealer grew suspicious of massive bulk orders. The seller hid an Apple AirTag inside a shipment. It tracked the package to an Amazon artificial intelligence facility known as VGT3. Worker interviews confirmed the routine. Tech teams destroy physical books to scan loose pages on fast sheet rollers. It cuts digitization costs to a fraction of normal methods.Amazon did not deny it. A spokesperson noted that the company buys books through regular channels to improve customer products, as reported by TechCrunch. But focusing only on Amazon misses the real crisis. This is a tech-wide gold rush. Court documents from a lawsuit against Anthropic, hosted on Writer Beware, revealed an identical project named “Project Panama”. The internal corporate goal was simple. They wanted to scan every book in the world by destroying them.This shift from scraping websites to shredding paper exposes a paradox. Silicon Valley is running out of data. Tech companies have exhausted the accessible internet. Even worse, the web is now full of AI-generated text. Training a new algorithm on data made by an old algorithm causes a fatal mathematical error called “model collapse”. The systems degrade quickly until they output total gibberish.To fix this, tech firms need clean, verified human thought. This has made books published before 2022 the most valuable commodity in technology. Older books are guaranteed to be free of AI contamination. They are the premium fuel for next-generation systems.To get this data, tech giants are exploiting a legal loophole. They hide behind a century-old framework called the First Sale Doctrine. Under this common-law rule, once you buy a physical copy of a book, you have the right to sell it, give it away, or destroy it.The First Sale Doctrine was built to protect individual property rights. It ensured that free citizens truly owned their items. AI giants have turned this principle upside down. Legal analysts at Goodwin Procter noted that federal Judge William Alsup signaled that scanning legally purchased physical books for AI training counts as “fair use” under copyright law in the Bartz v. Anthropic case.Tech giants use this logic to launder copyrighted material. As debated in outlets such as Reason magazine, by buying a paper book for a few dollars and shredding it, companies protect themselves against billions in legal liabilities.While courts may allow this, the cultural damage harms our civil society.First, it erases history. AI buyers target obscure, out-of-print local histories that have never been digitized. Often, only a few physical copies exist. When corporate buyers grab these volumes and run them through paper shredders, they erase physical human heritage. This is not preservation. It is corporate extraction.Second, it encloses the cultural commons. For centuries, human knowledge sat in libraries, universities, and independent bookshops. These are local spaces built for public learning. By destroying these texts, monopolies lock human literacy behind private paywalls. They blend the lifework of authors into black-box algorithms, then sell it back to the public as a subscription.Finally, this practice undermines free-market capitalism. Healthy markets require clear property lines, contract enforcement, and respect for labor. Independent bookstores are being turned into supply chain cogs for industrial data harvesting. Authors get a single, tiny royalty from the initial sale. Then, that book trains an engine built to automate their job. This is a predatory strip-mining of creative capital.The solution does not require a bigger federal bureaucracy. Instead, Washington must simply defend basic property rights.TRUMP’S FINRA FOR AI SOUNDS ODD, BUT IT MAY BE OUR BEST SHOTCongress should update property laws for the digital age. Industrial data extraction violates the spirit of the First Sale Doctrine. If a tech company wants to ingest a lifetime of human writing, it must negotiate voluntary, free-market licenses with creators.If we let tech firms slice apart our literary history to feed their algorithms, we face a dark turning point. This is a high-tech book burning, and the smoke is rising from Silicon Valley.Dr. Eric Wargotz is a political and policy analyst, writer and commentator, a former Senate nominee, and former elected president of the Queen Anne’s County Board of Commissioners in Maryland. He has held leadership roles across government and medicine. The views expressed here are entirely his own.