AI Companies Purchase Antiquarian Stock for Training Data

Antiquarians in Switzerland are experiencing a rise in sales of non-fiction and novels to AI companies, which use these stocks to generate training data. According to reports from Swiss Radio and Television (SRF), these are books that were previously considered slow sellers and are now being exported in large quantities to the USA. The demand for these books has surged in recent months. Antiquarians report that they regularly receive inquiries from AI companies specifically searching for certain titles. This development has led many dealers to strategically increase their inventories to meet the companies' demands.

One reason for the interest of AI companies in these books lies in the necessity to collect large amounts of text data. This data is crucial for training language models and other AI applications. The books provide a variety of writing styles and topics that are significant for the development of algorithms. Some antiquarians have already established special departments dedicated solely to sales to AI companies. These departments specialize in understanding the needs of the companies and offering suitable titles.

Dealers report that they are often specifically asked for certain genres or authors. The legal framework for selling books to AI companies is largely unclear. While some dealers ensure that they own the rights to the books, there are concerns regarding the use of the content for commercial purposes. Experts warn that the unregulated use of copyrighted materials could lead to legal conflicts. This development also impacts the book market as a whole.

Some publishers and authors express concerns that the use of their works by AI companies could occur without appropriate compensation or recognition. This could lead to a long-term decline in publications, as authors may be less willing to publish their works. The rising demand for books for AI training data has also resulted in an increase in prices for certain titles. Antiquarians report that prices for specific works deemed particularly valuable for AI training have doubled or even tripled. This has led some dealers to reconsider their pricing strategies.

The situation raises questions about the future of the antiquarian market. Dealers must adapt to keep pace with the changing demands of the AI industry. Some antiquarians are already experimenting with new business models to meet the demand while protecting their inventories. This development is part of a larger trend where AI companies increasingly rely on alternative data sources to train their models. The use of books is just one example of the creative approaches companies are taking to expand their databases.

According to estimates, up to 30% of the required training data could come from such sources. The discussion about the ethical implications of this practice is expected to intensify. Experts are calling for clear regulations to ensure that the rights of authors and publishers are upheld. A first step could be the introduction of licensing models that guarantee fair compensation for the use of content. The debate over the use of antiquarian stocks for AI training data is also being addressed at the political level. Some legislators have already begun to tackle the challenges arising from the use of copyrighted materials. A bill to regulate these practices could be introduced in the coming months.