Key facts
- AI companies are purchasing physical books in bulk.
- Books are destroyed after being scanned for training data.
- This practice has increased demand for obscure and rare titles.
- A US judge ruled that AI company scanning is protected by fair use.
- Anthropic is one of the AI companies reportedly involved.
- Concerns exist about the permanent loss of literary works.
- The practice is described by some as 'dystopian'.
Artificial intelligence companies are engaged in a practice of acquiring physical books in bulk, subsequently destroying them after scanning the content for training large language models. This surge in demand has reportedly driven up prices for obscure and rare titles. The practice has drawn criticism and raised concerns about the potential permanent loss of literary works, even as recent legal rulings have offered some protection to AI firms.
A US judge has ruled that the scanning process employed by AI companies, such as Anthropic, is protected under the doctrine of fair use. This legal precedent suggests that using copyrighted material for AI training may be permissible without explicit permission from copyright holders. Despite this legal backing, the methods used by these companies are being described as 'dystopian' by some observers due to the scale of book destruction and its impact on cultural heritage.
The core issue revolves around the tension between technological advancement and the preservation of physical media and intellectual property. As AI models require vast datasets to learn and improve, companies are seeking novel sources of information. The destruction of physical books, while efficient for data acquisition, poses a threat to the availability of these works in their original form and could impact collectors and libraries. The long-term implications for literary preservation and access remain a significant concern.