Japan Publishers Lose Trust After Nippan Sold Books for AI Training

Overview

Trust between Japanese publishers and Nippan Group Holdings is eroding rapidly following revelations that the distributor’s subsidiary sold books to American artificial intelligence developer Anthropic PBC without notifying copyright holders. The disclosure, unearthed in U.S. court documents, has ignited fierce criticism within the publishing industry over the unauthorized use of literary works for AI training.

Norio Yamamoto, president of the Japan Book Publishers Association, which represents approximately 380 member publishers, condemned the discrepancy between Nippan’s stated values and its actions. He emphasized that distributors have historically claimed to support publishing culture through close cooperation with publishers and bookstores, a promise that now appears broken.

"What they say and what they’re doing are different things." — Norio Yamamoto, President, Japan Book Publishers Association

This incident underscores the growing tension between traditional intellectual property rights and the insatiable data hunger of the AI sector. As legal frameworks lag behind technological advancements, publishers are left grappling with significant economic and ethical concerns regarding the exploitation of their creative works.

The Legal Exposure and Corporate Response

The controversy came to light through litigation filed in August 2024 by U.S. authors against Anthropic in the U.S. District Court for the Northern District of California. The lawsuit alleges copyright infringement, accusing the AI giant of digitizing books and acquiring data from pirated websites to train its models. Documents submitted as part of this case identified "Nippan Japan," believed to be a subsidiary of Nippan Group Holdings, as one of the suppliers from which Anthropic purchased books.

According to investigative reports, Anthropic was reportedly conducting a project involving the mass procurement of physical books. In this process, the company allegedly cut off the spines of numerous volumes to facilitate scanning of the pages, creating massive datasets for AI development. While the lawsuit focused on Anthropic’s practices, the involvement of a major Japanese distributor highlighted the global supply chain issues inherent in AI training data acquisition.

In response to inquiries from the publishers' association, Nippan acknowledged early this month that it had indeed sold books to Anthropic. However, the company maintained that it was unaware the purchases were intended for AI training purposes. A representative told The Yomiuri Shimbun that they would not comment on the specifics of individual transactions, citing privacy and commercial sensitivity.

Despite these assurances, multiple industry sources indicate that Nippan officials have begun visiting publishers to provide explanations. This shift from denial to direct engagement suggests that the distributor recognizes the severity of the reputational damage and the potential for long-term contractual fallout.

Financial Pressures and Industry Skepticism

The timing of the scandal has drawn particular ire, as Nippan Group is currently facing significant financial headwinds. The broader publishing industry in Japan has been in a prolonged slump, affecting Nippan’s core wholesale business. In fiscal 2025, the company recorded consolidated net sales of ¥342.7 billion but suffered an operating loss of ¥1.3 billion. More alarmingly, its wholesale segment alone incurred a loss of ¥3.6 billion, intensifying pressure to restructure operations.

These financial struggles have fueled skepticism among industry insiders. Some employees at Nippan have suggested that management may have been aware of the AI usage, driven by the need to diversify revenue streams during a period of declining traditional sales.

"This is the worst possible timing. This happened when we’re already being stretched to the limit due to the decline in the company’s business performance. I think management was aware the books were being used for AI training." — Nippan employee, speaking to The Yomiuri Shimbun

While Nippan claims ignorance of the end-use, the financial motive provides a plausible explanation for why such a large volume of books would be sold to a single foreign tech entity without standard publisher notifications. The perception that the distributor prioritized quick profits over professional ethics has further damaged the delicate relationship between wholesalers and publishers.

A Legal Gray Area in Japan

Japan’s current legal landscape creates a unique environment for AI developers, often described by critics as a "paradise for AI training." This characterization stems from Article 30-4 of Japan’s Copyright Law, added during a 2018 revision. This provision permits the use of text, illustrations, and other works for AI training without seeking permission from or providing compensation to copyright owners.

Consequently, there are no established judicial precedents in Japan declaring it illegal to purchase legally obtained books and scan their content for AI training purposes. Copyright lawyer Kensaku Fukui noted that the transactions involving new books sold by Nippan were likely legal under the current framework.

"As things stand, it’s quite unlikely that the Nippan transactions that sold new books were illegal. I think we can probably say this will become an issue to be ironed out in future contract talks between publishers and wholesalers." — Kensaku Fukui, Copyright Lawyer

However, Fukui also highlighted the stark inequity in the arrangement. The economic benefits accrued by AI development companies like Anthropic are disproportionately large compared to the minimal or nonexistent compensation received by Japanese publishers. This imbalance has sparked broader debates about intellectual exploitation and the need for equitable remuneration in the digital age.

In contrast, the European Union has taken a more regulatory approach. Enacting the first comprehensive AI law in 2024, the EU introduced limitations on the automatic analysis of text and illustrations by AI systems. Japan has yet to amend its Copyright Law to address these concerns, leaving publishers vulnerable to similar practices by both domestic and international AI firms.

Looking Ahead

The Nippan case serves as a catalyst for potential regulatory reform and stricter contractual enforcement within the Japanese publishing industry. Publishers are likely to demand greater transparency and compensation mechanisms for the use of their works in AI training datasets.

As the legal boundaries remain unclear, the immediate future will likely see intense negotiations between publishers and distributors. The industry is calling for a reevaluation of how AI data is sourced, emphasizing that while innovation is vital, it should not come at the expense of creators' rights and livelihoods. Until Japan updates its legal framework to match evolving global standards, tensions between the publishing world and the AI sector are expected to intensify.