An AirTag in an order of rare books revealed destructive scanning for AI training

For some time now, sellers of old and rare books have noticed a strange trend: anonymous customers were buying up massive volumes of books regardless of price. Journalists from 404 Media partnered with a bookseller who received an order for about a thousand books and hid an Apple AirTag tracking device inside one of them.

Knowledge of training data origins is gaining importance

The ability to prove where big players extract information for their LLMs is crucial. Precisely because of copyright and unauthorized use of content, lawsuits against corporations are increasing.

The tracking device pointed to an Amazon training facility

The shipment containing a thousand books and a hidden AirTag ended up at an Amazon warehouse labeled LAS8 in Las Vegas. Specifically, in the VGT3 department. Discussions among Amazon employees on online forums subsequently confirmed that this part of the building is exactly where the destructive scanning of massive volumes of text occurs.

Books destroyed for data

This is not careful archival work with gloves. The terminology of destructive scanning means that the spines of the books are chopped off so the pages can be fed into rapid machine scanners. The physical copy of the work is permanently destroyed during this process in exchange for the high-quality data that companies like Amazon then use to feed their language models.

Physical evidence connects anonymous purchases to Big Tech

Until now, the connection between massive book purchases and AI training was mostly speculation, partly involving Anthropic. Physical tracking has shown that these companies acquire high-quality copyrighted works through intermediaries and secondary market purchases. This revelation will serve as another strong argument for authors demanding compensation.

Lilith's verdict

Corporations have stopped vacuuming just the public web and are physically destroying culture on a massive scale to acquire high-quality data. AI training is no longer just about code; it has very tangible victims in the form of thousands of destroyed books.