TECH Signal 484
A 'bananas' order for 5000 obscure book titles fuels suspicion
↗
European booksellers are receiving large, automated-seeming orders for thousands of obscure books, suspected to be part of AI companies' efforts to acquire training data by scanning and destroying physical texts.
This reveals the physical supply chain behind AI training data acquisition, companies are buying and destroying physical books to digitize their contents. For engineers, it shows how far AI builders will go to acquire training data and the legal arbitrage being exploited by routing books through Europe to the US where 'fair use' permits scan-and-destroy practices.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
European booksellers are receiving large, suspicious orders for thousands of obscure books, suspected to be automated AI-driven purchases for training data acquisition.
Anthropic's 'Project Panama' was revealed in court documents to have spent 'tens of millions of dollars' acquiring and scanning millions of books for AI training, resulting in a $1.5 billion settlement with authors.
The orders appear designed to exploit legal differences between jurisdictions, books are collected in Europe then shipped to the US where 'fair use' policies permit scanning and destroying owned books.
THE CLUSTER