Amazon, once an online bookseller, is destroying rare books to train AI models
What happened
Amazon has reportedly been destroying rare physical books in its warehouses to convert their content for training large language models (LLMs). These rare books are valuable because they contain information not already available in the public internet data sets that AI models typically train on. The move marks a controversial shift for Amazon, which started as an online bookseller and now leverages these physical assets to fuel AI development.
Why it matters
Most AI training data comes from publicly available digital content, which means the novelty of the information is limited. Rare books often contain unique knowledge, deep historical texts, or niche subject matter that can improve AI understanding and creativity. By sacrificing physical copies to digitize and ingest these texts, Amazon gains a competitive advantage in training more comprehensive and nuanced language models. For operators and businesses, this signals a growing premium on high-quality, rare data sources that push AI capabilities beyond standard internet crawls. It also exposes a cost dynamic where access to exclusive datasets can overshadow basic data gathering.
What to watch next
The practice raises questions around cultural preservation versus commercial gain, especially as more companies seek rare content for AI training. Watch for industry pushback on data sourcing ethics and potential regulations on destroying valuable physical media. The approach could influence how AI builders source unique datasets, with operators needing to evaluate the trade-offs between data exclusivity and reputational risk. Additionally, monitor how Amazon’s AI offerings evolve with this data edge and whether competitors adopt similar tactics that impact rare book markets and libraries.
AI Quick Briefs Editorial Desk