Amid growing global controversy over artificial intelligence companies purchasing massive quantities of books for data training, it was confirmed on October 26 that a US firm supplying learning data to AI corporations recently recruited personnel to buy books in bulk in South Korea.
US Data Engineering Firm Recruits Book Sourcing Specialists in Korea
According to LinkedIn, the US-based company Innodata posted a job opening 46 months ago for an AI data companies buying books initiative, specifically seeking an ‘Academic Book Procurement and Logistics Manager’ with the location specified as ‘South Korea.’ The qualifications included 10 or more years of experience in used books, academic publishing, library acquisitions, or book distribution; a proven track record of procuring academic books in bulk across multiple countries; and strong networks with universities, libraries, wholesalers, distributors, and used book trading platforms. A total of 19 applicants were identified via LinkedIn, though it remains unconfirmed who was ultimately hired.
Innodata, a US Nasdaq-listed company, introduces itself in corporate briefing materials as a global data engineering enterprise. Specifically, it explains that its business involves supplying large-scale data required to build reliable AI systems. Considering the job postings, the company appears to purchase books globally, scan them, and convert them into data for AI training to provide to AI corporations. Academic books are reportedly the primary data preference for AI companies.
Global Expansion into Bulk Used Book Acquisition
Innodata posted similar recruitment notices targeting not only South Korea but also China, Indonesia, and the broader Asia-Pacific region around the same timeframe. Notably, the postings for Indonesia and the Asia-Pacific region set an initial target of ‘securing 100,000 to over 200,000 used academic books.’ The role involves securing books from universities, libraries, and inventory disposal companies and exporting them to designated countries. Experience in procuring used books in science, medicine, law, and technology was also listed as a preferred qualification.
The ‘Panama Project’ and Industry-Wide Book Scanning Trends
The US AI company Anthropic executed the ‘Panama Project’ starting in 2024, purchasing paper books in bulk, scanning them, discarding the physical books, and feeding the scanned data into AI models. A US court ruled in June of last year that this Panama Project did not constitute infringement, classifying it as ‘fair use.’ Consequently, an atmosphere of industrialization surrounding the procurement of books for AI learning is spreading primarily centered around the United States.
For instance, the startup Last Token, launched this February, states on its website that it engages in businesses such as ‘bulk book scanning’ and ‘collecting multi-language text from books and web sources.’ It has also promoted its capacity to scan up to 400,000 books per month. The CEO of this company shared recruitment news on LinkedIn, explaining that books are scanned and subsequently destroyed, describing the practice as ‘legal.’ LinkedIn has also featured job postings from global staffing firms seeking contract workers to slice books and digitize them using industrial scanners.
Meanwhile, Nippon Shuppan Hanbai (Nippan), which exported 122 tons of books to Anthropic, announced through the Nikkei newspaper on the 23rd that it will establish an internal trading ethics review committee in response to domestic criticism in Japan. In South Korea, Woongjin Booksen—which also exported books to Anthropic—distributed a statement to publishing organizations on the 21st declaring that it will further strengthen its book export management and review procedures.
