U.S. advocacy groups are urging the Federal Trade Commission (FTC) to investigate the practices of big tech companies that purchase books in bulk for artificial intelligence (AI) training, only to scan and then destroy the originals.
On August 21, the advocacy group Demand Progress Education Fund and 17 other organizations sent a joint letter to FTC Chair Andrew Ferguson and Commissioner Mark Meador, requesting an investigation into whether the book purchasing and destruction practices of AI companies constitute unfair competition, according to a report by Yonhap News on August 22.
The groups argue that the practice of buying large quantities of physical books, scanning them, and permanently destroying the originals blocks competitors from accessing the same materials.
They emphasized that this 'source blocking' makes it difficult for emerging AI startups to enter the market and called for the FTC to apply federal laws against unfair competition in its investigation.
In their joint letter, the organizations criticized AI companies for permanently destroying books in bulk, thereby restricting public access to these materials, stating, "Only the wealthiest existing companies are able to build high-quality AI models."
They also claimed that by eliminating original works, a future is being created where only certain companies hold exclusive rights to humanity's records. They argue that many of the books used for AI training were published before the rise of generative AI, making these materials finite and a scarce resource for humanity.
Reports indicate that some big tech companies have engaged in this method of acquiring AI training data.
Anthropic has built its AI training materials through a project known as the 'Panama Project,' where it purchased millions of books, cut off their spines, scanned them, and then recycled the destroyed books as paper waste.
Amazon is also reported to train its AI models using a similar approach. The Guardian recently reported instances of bulk purchases of used books from secondhand bookstores in North America, Europe, and Australia that appear to have been selected by machines.
IT media outlet 404 Media confirmed that a bookstore that received a large order attached tracking devices to 1,000 books, which were found to be headed to an Amazon scanning facility in Las Vegas.
This controversy extends beyond AI training and copyright issues to encompass fairness in competition among AI companies.
Previously, a U.S. federal court ruled that using legally purchased books for AI training constitutes 'fair use' and does not violate copyright law in a lawsuit brought by authors against Anthropic. However, advocacy groups argue that even if the destruction of books does not constitute copyright infringement, it may still amount to unfair competition.
Anthropic recently settled a copyright lawsuit with authors for $1.5 billion, reportedly the largest copyright settlement in U.S. history.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.