AI companies are burning books, advocates complain to FTC

AI firms are acquiring quantities of physical books for model training, subsequently destroying the originals. A coalition of 18 advocacy groups has petitioned the US Federal Trade Commission to investigate this practice. Details of the operation, known as Project Panama at Anthropic, emerged in court filings from a copyright lawsuit brought by authors whose works were used without authorization.

Anthropic dubbed the initiative "Project Panama" to avoid public awareness, as it states in an internal memo from 2024. Older books are deemed valuable for AI training since they remain unspoiled by AI-generated text, which has infiltrated recent written works. Destroying the books eliminates storage expenses. In certain cases, scan-and-destroy procedures may bolster fair use arguments.

However, the Internet Archive's attempt to utilize this legal defense was unsuccessful. Anthropic, as well as Amazon, have been involved in book scanning and discard operations, but neither company has commented on the allegations. This practice has become a public relations nightmare due to the destructive nature of book destruction and its connection to authoritarian regimes, as well as the widespread criticism of AI firms for appropriating public resources for private gain.

Intermediaries, such as ISBNdb, are also experiencing negative repercussions for facilitating the illicit acquisition of books. Civil society organizations are urging the FTC to probe the book buy-and-destroy activities, arguing that they impede competition and access to knowledge for both rival AI developers and the general public.