Civil Society Groups Ask FTC to Probe AI Book Destruction

Civil society coalition targets AI training practices

More than a dozen civil society organizations have formally asked the Federal Trade Commission to investigate AI companies that destroy physical books after using them to train large language models. The coalition — which includes Demand Progress Education Fund, the Consumer Federation of America, and the Institute for Local Self-Reliance — argues the practice may violate antitrust law.

The organizations contend that by destroying books after scanning them, AI companies are deliberately limiting the supply of training materials available to competitors. Each destroyed volume represents one less resource that rival firms or startups can use to develop competing AI systems.

Why it matters

This complaint reframes what many view as a cultural or ethical issue into a competition law question. If the FTC finds merit in the antitrust argument, it could establish legal precedent for regulating how AI companies acquire and dispose of training data — a dimension of AI governance that has received little regulatory attention despite its market implications.

Targeting pre-AI era content

According to the groups' letter to the FTC, AI companies are specifically seeking older books, particularly those published before 2022. These volumes are valuable because they are guaranteed not to contain AI-generated text and are more likely to have undergone rigorous editorial review.

The letter characterizes the practice as "hoard-and-destroy," arguing it functions as a structural mechanism to raise costs for competitors and deny startups access to essential source material. The groups frame it as "the latest escalation in a documented pattern of anticompetitive conduct designed to create an insurmountable systemic moat around AI incumbents."

Competition concerns over knowledge access

The antitrust angle highlights a tension in how AI companies position themselves. OpenAI CEO Sam Altman has previously described knowledge as a utility that could be sold "like electricity and water." But that framing becomes problematic when the companies building these knowledge utilities are simultaneously removing books — traditional repositories of human knowledge that remain widely accessible through libraries and used book markets — from circulation.

The civil society groups argue this practice could effectively lock competitors out of the AI marketplace by controlling access to high-quality training data. Unlike digital datasets that can be copied infinitely, physical books that are destroyed cannot be recovered or shared.

Regulatory path forward

The complaint represents an attempt to address AI training practices through existing antitrust frameworks rather than waiting for new AI-specific legislation. Whether the FTC will open a formal investigation remains to be seen, but the letter establishes a legal theory that could influence how regulators think about data acquisition strategies in the AI sector.

The details were first reported by Axios.