Anthropic is Burning Books. Every Book in the World, Actually.

Overview

An AI company, Anthropic, is accused of secretly scanning and physically destroying millions of real books—including rare volumes—as part of “Project Panama.” A planning memo allegedly describes the effort as “destructively scan all the books in the world.”

The video’s creator argues this is a modern, silent form of “burning books,” carried out not with flames, but through industrial equipment that:

• cuts out book spines,
• scans the pages,
• and sends the remaining physical books to disposal.

Framing via Three Documents

1. Company memo (unsealed via lawsuit)

• Anthropic allegedly planned the work while explicitly not wanting it to be publicly known.
• The video claims used books were bought in very large quantities and processed with cutters and scanning equipment, then destroyed.

2. Court order (San Francisco, June 20, 2025)

• A federal judge ruled the practice legal, relying on the legal logic: “one copy in, one copy out.”
• The video explains the judge’s distinction:
  • scanning while keeping the original could mean extra copies exist,
  • but destroying the physical book after scanning means the world still has effectively one extant copy—now in digital file form.
• The video notes Anthropic was fined $1.5 billion, but argues the fine was not for destructively scanning itself.
• Instead, it ties the broader case history to earlier alleged piracy—7 million books allegedly taken from pirate sites—suggesting the company later shifted to “legit” acquisitions followed by destructive scanning.

3. Broker/blog coverage (ISBNdb and ongoing activity)

• A broker allegedly sells bulk “AI training” scanning jobs under strict NDAs.
• The blog allegedly claims that even if “AI company destroys 2 million books” were public, it wouldn’t generate sympathy because buyers are deliberately obscured.
• The video also cites a used book dealer reporting a sharp increase in high-volume sales since April, including rare out-of-print titles.

Core Claim and Analysis

The video argues the destruction is strategically driven, not incidental:

• AI companies need scanned text.
• They also manage copyright risk by removing the physical book after scanning.

It further claims the most valuable targets are older books (pre-2022), because newer material is allegedly contaminated by AI-generated “slop.” The creator argues this creates a feedback loop:

• training on AI output
• that was itself trained on prior AI output

This is described as leading to “model collapse,” so companies supposedly seek “cleaner” older sources.

Impact Described

The creator emphasizes that for rare books, destructive scanning makes the final physical copies effectively unavailable:

• even if only a few copies exist publicly,
• the books’ text ends up in private servers,
• and models don’t “return” the original material as a readable book.

So the concern is not only legal questions around copyright, but the disappearance of cultural artifacts without meaningful public access.

Bottom-Line Framing

The video concludes that “Fahrenheit 451” got it wrong. Rather than a visible public crackdown, it depicts a quiet commercial pipeline:

• purchase orders,
• cutting/scanning hardware,
• and disposal,

that turns books into inaccessible training data, largely shielded by NDAs and corporate control.

Presenters / Contributors

• Corbin (host/creator), “Corbin Uncovers AI”