WASHINGTON, August 20 — A report by The Intercept alleges that Amazon has systematically destroyed rare and out-of-print books to source text for its artificial intelligence training, a practice reportedly initiated in 2022. The allegations, cited by internal documents and interviews with historians, describe the removal of books from libraries and private collections, including works by 19th-century authors and limited editions. Amazon has not publicly acknowledged the claims, though its AI division, Amazon AI, has used these texts to train models such as Alexa and Amazon Bedrock.
The destruction, as described by historians, threatens irreplaceable cultural artifacts. Many of the books are unique or in poor condition, making their loss irreversible.
Historians emphasized that the practice undermines efforts to preserve literary and historical knowledge, particularly for works that lack digital alternatives. The report notes that Amazon’s AI training datasets—comprising billions of words—may have incorporated text from these destroyed books, raising questions about the ethical boundaries of corporate data acquisition.
Corporate Practices Under Scrutiny
Internal documents obtained by The Intercept suggest Amazon’s approach was methodical, targeting books with high textual value but low commercial availability. This strategy, according to the report, aligns with a broader corporate calculus to maximize AI training efficiency while minimizing costs. Historians noted that the destruction of rare texts could skew AI models, potentially erasing nuanced historical perspectives or creating biases in language processing.
One historian added that the practice reflects a “fundamental disregard for the cultural and intellectual value of material that exists outside corporate profit margins.”
Ethical and Legal Implications
The issue also raises questions about the ethics of AI development. If Amazon’s models were trained on text sourced from destroyed books, the technology’s outputs could inadvertently perpetuate the erasure of certain histories or voices. This tension between innovation and preservation is not new, but the scale of Amazon’s alleged actions has intensified scrutiny.
The report does not detail whether other tech firms engage in similar practices, but it highlights Amazon as a focal point in a growing conversation about the moral costs of data-driven AI. Book lovers and scholars have expressed alarm at the potential loss of cultural heritage.
The destruction of out-of-print works, many of which are already scarce, could limit access to knowledge for future generations. The American Library Association has urged policymakers to address gaps in copyright law that enable such practices. Meanwhile, Amazon has not commented further, though the report notes the company’s AI division continues to expand, suggesting the practice may persist unless intervened upon.
The broader stakes of this controversy extend beyond books. As AI systems increasingly shape information ecosystems, the sources of their training data become a critical point of accountability.
The report frames Amazon’s actions as part of a geopolitical chessboard where corporate interests clash with cultural preservation. What this signals, as one historian put it, is a reckoning for how technology companies prioritize data acquisition over ethical considerations. The long-term consequences—whether for AI reliability, historical research, or public trust—remain to be seen.
