Towards a New Copyright Paradigm for Generative AI: Bridging Human-Machine Creativity Through Legal and Policy Reform

Daiwon Hyun, Sunho Park, Eutteum Kim; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2025, pp. 4487-4494

Generative AI challenges copyright frameworks for cultural heritage preservation. We analyze U.S. and EU approaches to AI-generated content through comparative legal analysis, revealing fundamental inadequacies in addressing machine learning's distributed agency. Recent 2025 rulings including Bartz v. Anthropic establish that AI training's legal status depends on transformativeness, data provenance, and market harm rather than categorical rules. We propose a five-principle harmonization framework: transparency through content provenance standards, protected research activities, tiered compensation for training and outputs, evidentiary requirements for creative control, and risk-based compliance. Analysis of recent policy developments including the U.S. Copyright Office's January 2025 guidance and EU's GPAI Code of Practice reveals emerging consensus around human authorship requirements. Our framework provides practical guidance for navigating divergent jurisdictional approaches while enabling international interoperability in AI-driven cultural preservation.