GENERATIVE AI AND COPYRIGHT LAW: A COMPARATIVE ANALYSIS OF OWNERSHIP RIGHTS IN CONTENT PRODUCED BY LARGE LANGUAGE MODELS

Abstract

The rapid proliferation of Large Language Models (LLMs) and multi-modal generative artificial intelligence systems has fundamentally disrupted international intellectual property law, challenging the foundational assumption that expressive creation is an exclusively human endeavour. This review article presents a comprehensive comparative analysis of ownership rights in AI-generated and AI-assisted content across four major jurisdictions: the United States, China, the United Kingdom, and the European Union. Drawing upon recent judicial rulings, administrative guidance, and statutory frameworks, the analysis examines how divergent philosophical foundations ranging from strict anthropocentric human authorship requirements to pragmatic economic incentive models are fragmenting the global legal landscape. The United States maintains a rigid human authorship doctrine, denying copyright protection to autonomous machine outputs while extending limited protection to human-authored elements in hybrid works. China has emerged as a notable exception, with courts vesting copyright in users who demonstrate iterative creative choice through prompt engineering. The United Kingdom's Computer-Generated Works framework statutorily allocates ownership to the person making necessary arrangements, while the European Union rejects copyrightability for pure AI outputs lacking human intellectual creation. The article further explores the dual risk paradigm confronting enterprises: downstream asset unprotectability and upstream infringement liability. It proposes corporate compliance protocols, including prompt logging, source code scanning, human creative intervention, and upstream data verification, to mitigate legal exposure. The analysis concludes that regulatory arbitrage opportunities and cross-border enforcement challenges necessitate a harmonised international approach to AI-generated content governance.