Urgent.News

What's breaking now, across thousands of outlets.

Editions

AI

AI companies are burning books, advocates complain to FTC

Fahrenheit 203, the temperature GPUs stop gorging on literature

AI companies are burning books, advocates complain to FTC

Dozens of advocacy groups have approached the US Federal Trade Commission in a bid to probe the practice of AI firms purchasing physical books for model training, only to subsequently destroy the originals. This disturbing revelation surfaced in the courtroom during a copyright dispute brought forth by authors whose books were used for AI training without their consent.

The AI companies involved include Anthropic and Amazon, both of which declined to comment on the matter. The practice, known as 'scan-and-destroy', was unveiled in a 2024 internal memo by Anthropic, which dubbed the operation 'Project Panama'. The memo explained that the company aimed to scan all books on the planet and then destroy the originals, a fact that was strictly confidential and not to be discussed outside the company.

Older books, devoid of AI-generated content, are deemed valuable for AI training as they are 'unpolluted', and eliminating the physical copies after scanning helps avoid storage expenses. While such actions could potentially support fair use claims, this practice did not favor the Internet Archive in the Bartz case. This method of acquiring and destroying books is not limited to Anthropic; a recent report suggests that Amazon is also involved in book scanning and shredding.

The revelation has sparked public outrage, as it aligns with the negative perception of AI companies for exploiting public resources for private gain. Additionally, the destruction of books, reminiscent of authoritarian regimes, has tarnished the image of these AI companies. Intermediaries involved in the acquisition process, such as ISBNdb, have distanced themselves from the practice.

In response to these revelations, civil society groups are urging the FTC to investigate the scan-and-destroy operations, arguing that they hinder competition and monopolize knowledge. The groups contend that these operations are anticompetitive and could lead to a future where only the wealthiest companies can develop advanced AI models, effectively controlling humanity's written works.

They ask the FTC to answer whether the destroyed physical books are the last or among the last surviving copies of a given work, given the companies' reluctance to disclose their training data.

Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at theregister.com →

More in AI

AI Is Opening New Revenue Streams for Automakers

Artificial intelligence (AI) could give automakers something they have spent years trying to build: a way to keep making money from a vehicle long after it leaves the dealership. Ford and General Motors (GM) are both betting that an AI assistant tied to a vehicle’s data can keep that relationship, and the revenue, going long […] The post…

More from Friday 21 August →