The U.S. Federal Trade Commission is being urged to investigate whether some AI companies obtained training data by purchasing physical books, scanning them and destroying the originals.
According to a public letter obtained by Axios, more than a dozen civil society organizations asked the FTC to use its regulatory authority to review what they called a 「destructive new way of acquiring data」 used by large AI companies.
Earlier, The Washington Post reported, citing court filings, that Anthropic spent millions of dollars buying books, removing their bindings and scanning the pages to train Claude. Google, Microsoft and OpenAI have also faced similar copyright lawsuits.
The groups want the FTC to determine whether this practice could qualify as an unfair method of competition. In their view, acquiring and destroying physical books could hollow out key data resources in the market, especially if some rare titles disappear permanently while digital companies control the last remaining physical copies.
The letter also warned that this 「hoard and destroy」 approach could raise data acquisition costs for rivals and cut off access to source materials that startup AI companies rely on to train models, widening the competitive barriers around the largest firms.
The organizations are not asking the FTC to limit AI model training itself. Their request is focused on the destruction of existing works and whether regulators should intervene before large AI companies use the practice to build market advantages.
The public letter said the conduct was not merely a data purchasing strategy, but could become another structural way for major AI firms to build what it described as a 「systemic moat」 that would be difficult to cross.
Under the Trump administration, the FTC has sought to preserve a relatively business-friendly regulatory environment for U.S. companies while also continuing to signal concern about market competition and the monopoly risks tied to large technology companies.