Physical Ingestion Arbitrage and the Betrayal of Distribution Fiduciaries: Unconsented Mass Book Scanning for Generative AI and the Destruction of Publishing Sovereignty

The governance of generative artificial intelligence training datasets has historically focused on digital web crawling and online opt-out mechanisms. However, the revelation that Japan's primary book distribution oligopoly covertly executed massive, unconsented wholesale deliveries of physical printed volumes to American foundation model developer Anthropic marks an alarming paradigm shift: Physical Ingestion Arbitrage. To circumvent digital web barriers, terms of service restrictions, and European machine-readable opt-outs, frontier artificial intelligence corporations have turned toward industrial-scale procurement of physical books, physically slicing bindings, executing high-throughput optical character recognition, and assimilating entire national publishing heritages into proprietary latent weight spaces without authorial knowledge or equitable remuneration. This monograph analyzes the tri-fold institutional catastrophe precipitated by this practice. First, we examine the structural breach of fiduciary trust committed by wholesale distribution intermediaries, who transformed their role as cultural infrastructure into speculative brokers liquidating creator assets to foreign tech monopolies. Second, we formalize the legal arbitrage exploiting the asymmetry between Japan's permissive copyright exceptions and foreign commercial deployment, proving that physical-to-digital transcoding constitutes a willful infringement under international private law that bypasses fair-use protections. Third, we establish the doctrine of Somatic and Physical-Layer Rights Reservation, demonstrating why defensive publication via CERN Zenodo, cryptographic watermarking, and strict physical No-Learn covenants are indispensable to safeguard cultural sovereignty and prevent the irreversible extraction of human literature into enclosed artificial intelligence monopolies.