Japan’s pro-AI training ruling contrasts with EU transparency mandates and U.S. legal uncertainty, creating jurisdictional competition in the $67.8B AI training market.
A June 11 decision by Japan’s Copyright Committee declaring AI training exempt from copyright infringement has intensified global debates about AI development ethics, coming three days before the U.S. Supreme Court declined to review a pivotal AI copyright case. This regulatory split emerges as the EU finalizes mandatory training data disclosure rules, creating distinct jurisdictional approaches to the AI industry’s core controversy.
Japan’s Strategic Copyright Exemption
Japan’s Agency for Cultural Affairs confirmed on 11 June 2024 that AI developers can use copyrighted materials without licenses for training purposes, a policy the government says positions Japan as ‘the global AI development hub’ (Ministry of Internal Affairs statement). This follows protests from 23 creative industry groups who argue the policy enables ‘industrial-scale content laundering’ (Japan Artists Guild).
U.S. Legal Limbo Persists
The U.S. Supreme Court’s 14 June decision to deny certiorari in Andersen v. Stability AI leaves intact a Ninth Circuit ruling that compared AI training to ‘a student studying copyrighted textbooks’ – a precedent tech firms hail as fair use validation. However, 138 members of Congress recently co-sponsored the Generative AI Disclosure Act requiring model transparency.
Europe’s Transparency Compromise
The EU AI Act’s final draft (released 12 June) requires companies to document copyrighted training data but stops short of mandating licenses. ‘This is about traceability, not prohibition,’ explained EU digital commissioner Margrethe Vestager during a Brussels press conference. The policy aligns with OECD’s 13 June recommendation for voluntary transparency standards.
Industry Negotiations Intensify
The New York Times and Axel Springer are negotiating content deals with OpenAI, confirmed by insider reports on 10 June. These follow Shutterstock’s established AI licensing program that has paid $94 million to contributors since 2022 – a model some consider preferable to legislation.
Historical Precedents in Digital Copyright
The current debate echoes early 2000s battles over music streaming. When YouTube launched in 2005, it initially argued its platform constituted fair use before establishing Content ID and revenue sharing. Similarly, Google’s book scanning project survived a 13-year legal challenge, with courts ruling in 2016 that digitization for search qualified as transformative use.
Technological Shifts vs Legal Frameworks
Legal scholar Pamela Samuelson notes: ‘Copyright law always lags behind disruptive technologies – we saw it with photocopiers, VCRs, and web indexing. The AI training question might ultimately require new statutory definitions of machine learning’s technical processes.’ The 2024 OECD report emphasizes that 73% of generative AI training data comes from sources created after 2000, complicating fair use assessments designed for human learning analogs.
