ISBNdb promised AI firms raw book data — then suddenly deleted it
isbndb removed a page offering to scan and sell printed books to ai labs after media scrutiny, revealing the high demand for 'clean' pre-2022 data to prevent model collapse.
isbndb, a book database provider, deleted a landing page and removed articles offering 'printed books sourcing for your ai llms dataset needs' following an investigation by 404 media. the company claimed the page was merely a test of market interest and that it had never scanned or sold books, despite previously citing the 'optics problem' of destroying books during the process. the deleted materials specifically highlighted the value of pre-2022 printed books as 'slop'-free training data to prevent model collapse and suggested that buying used books bypasses author royalties.