AI Firms Accused of Destroying Books to Shut Out Rivals

More than a dozen civil society groups have called on the US Federal Trade Commission (FTC) to investigate dominant AI companies for buying, scanning and then physically destroying books used to train their models, Axios reports.

If the commission takes up the request, the dispute over AI training data could expand from copyright into competition law, with regulators examining whether dominant companies are eliminating resources that rivals need to compete.

Anthropic reportedly spent millions of dollars acquiring books and physically removing their spines so the pages could be scanned and fed into its Claude model. Google, Microsoft and OpenAI have faced similar copyright lawsuits, according to Axios.

The groups argue that buying up and destroying physical copies could “starve the market” of important source material, raise costs for smaller competitors and make it harder for them to compete. They warn that rare books could in some cases disappear altogether if AI companies acquire and destroy the last remaining copies.

The groups are not asking the FTC to restrict the training of AI models. Instead, they want it to investigate whether the destruction of existing works constitutes an unfair method of competition and, if necessary, intervene before the practice helps entrenched companies strengthen their market dominance.

(max)