BotBeat
...
← Back

> ▌

AnthropicAnthropic
INDUSTRY REPORTAnthropic2026-08-02

Australian Booksellers Raise Alarm Over Destruction of Rare Titles to Feed AI

Key Takeaways

  • ▸AI companies, including Anthropic, are engaging in 'destructive scanning'—purchasing physical books, slicing off spines for easier scanning, then pulping the remains to acquire training data
  • ▸Australian booksellers discovered they were caught in the AI supply chain after receiving suspicious bulk orders from Zoom Books, a Canadian intermediary facilitating the data acquisition
  • ▸A US court ruled destructive scanning lawful as 'transformative use,' but the practice is now understood to be widespread across multiple AI companies seeking training data
Source:
Hacker Newshttps://www.theguardian.com/technology/2026/aug/02/australian-book-sellers-alarm-destruction-rare-titles-ai-supply-chain↗

Summary

Australian secondhand booksellers are expressing serious concerns after discovering they may have been caught in an AI supply chain that destroys rare books. The alarming practice, known as 'destructive scanning,' involves AI companies purchasing physical books, slicing off their spines for efficient scanning, then pulping the remains. The discovery came after multiple booksellers noticed suspicious bulk orders from Zoom Books, a Canadian intermediary, through major secondhand marketplaces like AbeBooks and Biblio—orders that were price-insensitive and completely random in their selection of niche titles.

The practice came to light through a copyright lawsuit against Anthropic brought by book authors. Unsealed court documents revealed that Anthropic had engaged in destructive scanning and initially sought to keep the practice secret. A US judge ruled the practice lawful under copyright law as 'transformative use,' but concerns about its widespread adoption across the AI industry persist. The revelations have sparked alarm among rare booksellers who view their inventory as more than merchandise—they are cultural artifacts with historical significance, including 400-year-old documents recording Spanish Inquisition events.

Vendors like Tim White of Books for Cooks and Delfina Manor of Good Reading Secondhand Books report that these orders disrupt their operations and threaten irreplaceable items. For rare booksellers, the destruction of one-of-a-kind books and historical documents to feed AI model training represents something deeply troubling, raising urgent questions about what cultural heritage gets preserved in the age of artificial intelligence.

  • Rare booksellers are deeply concerned that irreplaceable cultural artifacts and unique historical documents are being destroyed without regard to their significance beyond their text content

Editorial Opinion

The conflict between AI companies' insatiable hunger for training data and society's obligation to preserve cultural heritage reveals a troubling asymmetry in our technological priorities. While courts have deemed destructive scanning legally acceptable, the practice raises uncomfortable questions about what we value and what we discard in pursuit of algorithmic advancement. The destruction of rare books and one-of-a-kind historical documents—items that may be the only surviving copies of their kind—for the sake of training data feels like a profoundly shortsighted trade-off. As AI becomes increasingly central to our infrastructure, we need stronger safeguards to protect irreplaceable cultural artifacts from being consumed by the machinery of machine learning.

Generative AIMLOps & InfrastructureMarket TrendsEthics & Bias

More from Anthropic

AnthropicAnthropic
RESEARCH

Anthropic's Opus 5 Cuts Prompt Injection Success Rate to 2%, Far Outpacing Competitors

2026-08-02
AnthropicAnthropic
INDUSTRY REPORT

The $5K Tell: How Anthropic's Claude Powers AI Vendor Pricing Strategies That Hide True Costs

2026-08-02
AnthropicAnthropic
POLICY & REGULATION

Global Nobel Laureates Issue Rome Declaration Calling for Coordinated AI Slowdown and Safety Measures

2026-08-02

Comments

Suggested

Academic ResearchAcademic Research
RESEARCH

Beagle Framework Brings GPU Acceleration to Symbolic Regression with Significant Performance Gains

2026-08-02
European CommissionEuropean Commission
POLICY & REGULATION

EU AI Act Takes Effect: Companies Must Disclose AI Use and Comply with Risk-Based Framework

2026-08-02
Google / AlphabetGoogle / Alphabet
INDUSTRY REPORT

Reddit and Major Publishers Challenge Google's AI Overviews as Traffic Impact Spreads

2026-08-02
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us