The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
AI companies, including Anthropic and others, are purchasing and scanning large quantities of old, rare, and out-of-print books from second-hand bookstores worldwide to train language models. After digitization, the physical books are destroyed, resulting in loss of cultural heritage and public access to knowledge.[AI generated]
Why's our monitor labelling this an incident or hazard?
The event involves AI systems (large language models) that require vast amounts of human-written text data. The companies' automated purchasing and destructive scanning of physical books to feed AI training datasets directly cause harm by destroying physical books, which are part of cultural heritage and community knowledge. This destruction is legally mandated but results in loss of property and cultural harm. The harm is materialized and directly linked to the AI system's development and use. Hence, this qualifies as an AI Incident due to harm to communities and property caused by the AI system's use and development.[AI generated]