Is AI Training Destroying Physical Books? A Neglected Cultural Controversy

HN post asks whether physical books purchased for AI training are being cut and destroyed, revealing the material costs of the corpus supply chain.
The article revolves around a Hacker News post with only 12 comments, whose title asks "Are we just going to ignore that books are being destroyed for AI?" The author notes the controversy may involve two scenarios: destructive scanning that cuts off book spines to improve scanning efficiency, and direct destruction of paper originals after digitization to save storage costs. The article emphasizes that mainstream AI ethics discussions focus on copyright, bias, and energy consumption, but rarely touch on the material sources of training data, and that a destroyed book means an irreplaceable physical copy disappears. Since the material consists only of the post title and screenshot, the specific companies, scale, and whether library collections are involved cannot be confirmed, making the topic more of a question to be verified. The author believes that in the context of a lack of transparency in the AI corpus supply chain, this question itself has value.
Introduction: An Overlooked Problem
A short post on Hacker News sparked discussion: "Are we just going to ignore that books are being destroyed for AI?" The poster provided no additional links or evidence, but the title itself is sharp enough—it points to a rarely discussed aspect of large language model training: the physical consumption of paper books.

This isn't the old debate about copyright or data privacy, but a more concrete, material question: when tech companies purchase physical books at scale to obtain high-quality training corpora, where do these books ultimately end up?
The Core of the Controversy: How Books Are "Sacrificed for AI"
Scan and Destroy? Two Possible Scenarios
The post title hints at scenarios that may involve two possibilities. One is destructive scanning after large-scale procurement—to improve scanning efficiency and image quality, some digitization processes simply cut off the spine, making the book impossible to restore. Another more extreme speculation is that some corpus processing pipelines directly destroy the paper originals after digitization to save on storage costs.
Neither practice is unprecedented in the publishing industry. Library digitization projects have long faced tension between "preserving the original" and "improving efficiency." But when the actor shifts from public institutions to commercial AI companies, the nature of the problem changes: carriers of public memory are being consumed for private commercial purposes.
Destructive scanning (also known as "destructive scanning") is a professional term in the digitization industry, referring to the process of directly cutting off the spine of a book and disassembling it into individual pages for high-speed scanning, in order to obtain flat pages and improve scanning speed and image quality. This practice is not uncommon in archival digitization and out-of-print book electronization, because traditional flatbed scanning requires turning pages one by one, which is extremely inefficient and prone to page curvature and shadows. The cost of destructive scanning is that the book cannot be restored and can only exist as loose pages or digital files. For commercial AI companies, facing tens of thousands of purchased books, destructive scanning has clear advantages in cost and time, but it also means that every scanned book undergoes an irreversible physical consumption.
Why This Issue Deserves Attention
Although the HN comment section has only 12 discussions, the word "ignore" in the title precisely captures the current situation—mainstream AI ethics discussions focus on copyright, bias, and energy consumption, but rarely touch on the material sources of training data. A book that has been cut and destroyed means an irreplaceable physical copy has disappeared. Even if the content has been digitized, the paper, binding, version information, and annotation traces all vanish with it.
Version information and annotation traces are unique information carried by books as physical objects. Different versions of the same book may have subtle differences in typesetting, preface, errata, and illustrations, and these differences are valuable for textual criticism and bibliography. Annotations, notes, bookplates, and signatures left by readers in margins constitute the "social life history" of books and are first-hand materials for studying reading history and knowledge dissemination. Digitization typically captures only the main text images or OCR text, and these physically attached additional information is often filtered out during scanning. Therefore, even if the content of a destroyed book has been digitized, its integrity as a cultural artifact still suffers irreversible loss.
Information Gaps from Multiple Perspectives
It must be honestly stated that the current material only includes the title and screenshot of an HN post with 12 points and 12 comments, with no detailed reporting, no company response, and no specific cases. This means the following cannot be confirmed:
- Which specific companies are doing this
- How large the scale of destruction is
- Whether library collections or the second-hand book market are involved
Therefore, this topic is currently more of a "question raised" than a "confirmed scandal." But it is precisely this lack of information that is worth pondering: if large-scale destruction is indeed happening, why hasn't systematic investigative reporting formed?
Blind Spots in the AI Corpus Supply Chain
Large model training requires massive amounts of high-quality text. As public web data gradually dries up, book scanning has become an important source. From procurement, transportation, destructive scanning to corpus cleaning, every link in this supply chain is rarely publicly disclosed. The controversy over book destruction reminds us: the "data hunger" of the AI industry is producing chain reactions in the physical world, and these reactions often remain outside regulation and public view.
The consensus emerging from the HN discussion is that this issue should not be ignored. The disagreement lies in severity—some see it as resource waste, some worry about cultural heritage damage, and others believe more evidence is needed to judge.
Public web data depletion is a key background in recent AI corpus discussions. As large model parameter scales expand, high-quality public text (such as Wikipedia, news sites, forum posts) is gradually being "exhausted," and more and more websites are restricting crawler access through robots.txt or paywalls. Research organization Epoch AI has estimated that high-quality public text data may approach depletion around 2026. This pressure directly pushes tech companies toward non-public or semi-public corpus sources such as book scanning, academic papers, and licensed content. Books have thus transformed from "cultural heritage" into "strategic resources," and their procurement, processing, and disposal methods have entered a track dominated by commercial logic—this is the structural root of the book destruction controversy.
Conclusion
A post with 12 upvotes is not enough to settle the matter, but the question it raises stands: when AI swallows humanity's knowledge carriers, should we ask about the fate of those pages? Before the corpus supply chain becomes transparent, this question itself has value.
Related articles

AI Writes Your Code—Do You Still Need to Learn Git? Five Reasons
AI coding agents handle Git operations automatically—is it still necessary to learn Git? Five reasons: team communication, reviewing agent PRs, monitoring code changes, fallback when AI goes down, and the joy of learning—plus how the focus of learning Git is shifting in the agent era.

Anthropic Reportedly Prices IPO at $83 Per Share, Valuing Company at $1.5 Trillion
Anthropic reportedly prices IPO at $83 per share, above the expected range, with a fully diluted valuation of $1.5 trillion. This article analyzes the signal significance of the pricing and its impact on the AI industry's financing environment.

Open-Source Code Hosting Platform Codeberg Suffers Outage, Sparking Debate in Developer Community
Open-source code hosting platform Codeberg.org suffered an outage, sparking Hacker News debate over nonprofit platform stability and open-source infrastructure dependency risks.