AI Companies Accused of Destroying Books for Model Training
Advocacy groups have urged the FTC to investigate whether leading AI companies are buying, digitizing, and destroying books to train large language models in an anticompetitive practice. The practice could reduce public access to source material and disadvantage competing AI developers, the groups argue.
What happened
More than a dozen public interest and consumer advocacy groups, including the Demand Progress Education Fund, Consumer Federation of America, and Institute for Local Self-Reliance, urged the FTC on Friday to investigate leading AI developers for allegedly buying, scanning and destroying books used to train their large language models.
The groups described the practice as 'hoard-and-destroy,' alleging that AI companies buy books in bulk, digitize the content to train their AI apps, and then destroy the physical copies, with some destroyed books being among the last surviving copies of original works.
Reported impact
The groups argue that such practices could make it harder for the public and other AI companies to access essential source material, and could engineer a future where only the wealthiest incumbents can build high-quality AI models while operating as the sole holders of humanity's written works.
A separate investigation by digital publisher 404 Media reported that Amazon is buying books in bulk, scanning them to train its AI tools and then destroying them.
Index analysis
The groups allege that destroying books could constitute an unfair method of competition under Section 5 of the FTC Act, which bars unfair or deceptive acts or practices in commerce.
The groups cite an August 2024 copyright lawsuit brought by writer Andrea Bartz and others against Anthropic, though in 2025 a judge ruled that Anthropic's use of legally purchased books to train Claude did not violate copyright law.
Governance controls
The groups are asking the FTC to determine the scale of the practice and how many of the destroyed books are the last remaining copies of the underlying works.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.