Global Divergence in AI Copyright Liability: A Comparative Analysis of Fair Use, Text and Data Mining, and Fair Dealing in the US, EU, and Caribbean (2025),
Global Divergence in AI Copyright Liability: A Comparative Analysis of Fair Use, Text and Data Mining, and Fair Dealing in the US, EU, and Caribbean (2025), By. Dr. Abiola Inniss Ph.D. LLM, ACIarb Abstract This article analyzes the diverging legal frameworks governing AI training and copyright in the United States, European Union, and Caribbean as of late 2025. In the United States , the judiciary has established a "conditional fair use" doctrine ( Bartz v. Anthropic , Kadrey v. Meta ), where training is transformative but liability arises from illicit data sources ("shadow libraries") or market substitution. The European Union enforces a statutory compliance regime under the AI Act, permitting text and data mining (TDM) only where rights holders have not exercised machine-readable opt-outs (e.g., C2PA). In the Caribbean , notably Barbados, legislative reforms prioritize creator sovereignty, rejecting broad TDM exceptions in favor of collective lice...