As artificial intelligence (AI) companies strive to enhance their chatbots, a controversial practice has surfaced involving the bulk purchase and destruction of millions of printed books. These physical books are being scanned page-by-page to create extensive digital datasets for training AI models.

Destructive Scanning and Its Implications

Known as destructive scanning, this process involves cutting off book spines, digitizing the content, and then recycling or discarding the physical copies. Court documents from ongoing litigation reveal companies like Anthropic have acquired books through third-party vendors to avoid direct publisher purchases, aiming to "destructively scan all the books in the world" while maintaining secrecy.

Anthropic defends the practice as fair use, citing the legal purchase of physical copies before digitization. However, this has sparked debates over copyright, the fate of printed literature, and ethical concerns about data sourcing.

Why Printed Books Matter in AI Training

Despite the vast amount of online content, much of it is now AI-generated, which risks degrading future models through repeated training on synthetic data. Printed books, by contrast, provide carefully edited, human-written text predating the AI era, making them invaluable for improving AI language quality. This includes classic literature, scientific journals, and academic works.

Potential Impact on Kenya's Literary Heritage

While the bulk of these acquisitions have occurred in the US and Europe, experts warn Kenya’s rich archives and rare publications could become targets for similar practices. Kenya's collections include indigenous knowledge, colonial-era texts, Kiswahili dictionaries, and out-of-print books that have yet to be digitized.

  • Should Kenya accelerate digitization efforts to safeguard its heritage?
  • Could local publishers negotiate licensing deals to protect works from unregulated export?
  • How will copyright laws adapt to the transformation of physical books into machine-readable data?

This trend paradoxically revives interest in printed books at a time when digital formats were expected to dominate. Rare and specialized publications may gain new value as AI seeks diverse, high-quality sources.

Legal Battles and the Future of AI Development

Authors and publishers worldwide are challenging AI firms over unauthorized use of copyrighted materials. Courts are grappling with balancing innovation and intellectual property rights, a dynamic that will influence AI's trajectory globally.

For Kenya, the stakes involve both opportunity and caution. The country’s unique cultural and academic resources could play a strategic role in AI’s evolution, provided they are preserved and managed thoughtfully.

Ultimately, the AI race is not just about technology but also about control over knowledge—the very foundations that will shape machines of the future.