Background & Context§
The rapid advancement of large language models (LLMs) has created an insatiable appetite for training data. While most AI firms rely on publicly available web text, a growing trend is the acquisition of proprietary, unpublished, or pre-digital content to gain a competitive edge. In a startling revelation, Anna's Archive—the world's largest shadow library—has exposed a secretive practice: AI companies purchasing and destroying physical books to ensure they are the only entities with digital copies. This news matters because it highlights a critical ethical crisis at the intersection of AI development and cultural preservation. As AI models become more powerful, the very source material that powers them is being systematically erased from the public sphere, raising urgent questions about access to knowledge and the future of human culture.
The News: What Happened Exactly§
According to a guest post on Anna's Archive blog (dated 2026-08-05), several AI companies have been acquiring large quantities of secondhand books through intermediaries, scanning them, and then destroying the physical copies. The primary goal is to obtain training data "untouched by machines"—content created before 2022, which is largely absent from the internet and therefore considered unique and valuable. This practice effectively removes these works from the public domain, creating a monopoly over knowledge.
Most notably, Anthropic's "Project Panama" was exposed in a $1.5 billion copyright settlement. Launched in early 2024, this highly confidential project involved spending tens of millions of dollars to purchase millions of paper books, scan them to train its Claude LLM, and then destroy the originals. The post describes this as "legally permissible" but an "extremely serious crime against humanity." The ethical outrage stems from the fact that after scanning, the AI company becomes the sole possessor of digital copies, locking away human knowledge on private servers forever. This is not just a copyright issue; it's a deliberate act of cultural vandalism.
The rationale behind destroying physical books is clear: by eliminating all other copies, the AI company ensures that no competitor or public entity can access the same content. This monopolization extends beyond mere data—it grants the company exclusive power over the knowledge itself. The post highlights a paradox: while AI companies promise to "make human knowledge accessible," they are dismantling the most durable carriers of that knowledge. The public may gain more intelligent AI assistants, but at the cost of losing access to the very resources those assistants are trained on.
Furthermore, the post reveals a larger existential threat. Since the beginning of 2025, AI-generated content has accounted for more than half of newly published internet content. If AI continues to absorb every last sentence written by humans on paper, and then generates new content based on that, the internet will become a closed loop of AI's own words. In such a world, preserving human civilization's authentic voice becomes nearly impossible. To combat this, Anna's Archive is urgently appealing to volunteers worldwide to scan and upload books, journal articles, newspapers, magazines, ancient texts, and rare materials from libraries and archives before they disappear. The call is simple: if every person scans just one book, and there are 10 million volunteers, we can preserve 10 million pieces of invaluable human heritage.
Historical Parallels & Similar Incidents§
This alarming practice is not without precedent. In 2004, Google launched the Google Books Library Project, an ambitious initiative to scan millions of books from major university libraries. The project aimed to make books searchable online, but it faced a massive copyright lawsuit from the Authors Guild and the Association of American Publishers. In 2013, a federal judge ruled in favor of Google, declaring that scanning books for indexing and search constituted "fair use" because it provided transformative value. However, unlike the current destruction, Google preserved the physical books and made snippets available to users. The intent was to increase access, not diminish it.
Another parallel comes from the world of pharma and academia, where certain researchers have been known to "plagiarize" or hoard data. More directly, in 2019, the U.S. government shutdown led to the National Archives and Records Administration (NARA) temporarily halting digital preservation efforts, but no destruction occurred. The closest analogy might be the historical burning of the Library of Alexandria, where a vast repository of human knowledge was lost forever due to conflict and neglect. While the motives differ—ancient conquerors versus corporate greed—the outcome is the same: permanent loss of cultural patrimony.
What makes the current situation more insidious is its deliberate and covert execution. Unlike Google Books, which openly partnered with libraries, these AI companies operate through intermediaries, destroying the physical artifacts to create artificial scarcity. This is a stark contrast to the open-access movement, which advocates for the democratization of knowledge. The lesson from the Google Books case is that legal battles can shape ethical boundaries. However, in the absence of clear regulation, companies are exploiting loopholes to prioritize profit over preservation. The precedent set by Google might have established a legal framework for scanning, but it also inadvertently gave a green light to more destructive practices.
Moreover, the reactive nature of the AI industry—driven by "move fast and break things"—means that ethical considerations often take a backseat to competitive advantage. The Project Panama leak underscores the need for oversight and transparency. As shadow libraries like Anna's Archive step up to fill the gap, they become modern-day guardians of knowledge, reminiscent of the monks who painstakingly copied manuscripts during the Dark Ages. The current effort is a race against time, involving volunteers from every corner of the world. The outcome will determine whether future generations inherit a rich, diverse intellectual heritage or a sterilized, corporately controlled version of history.
In conclusion, the destruction of physical books by AI companies is a pivotal moment in the history of information. It forces us to confront the unintended consequences of AI innovation and the fragility of our collective knowledge. The response from Anna's Archive and the global community may well define the legacy of the digital age.