AUTHORSHIP UNDER ZERO-EDIT POST-GENERATION MODIFICATION A Process-Based Framework for Examining Human Agency, Creative Control, and Authorship When Generative AI Performs the Visual Execution

Generative artificial intelligence has created a difficult question for contemporary artistic practice: when a human artist conceives an artwork, directs a generative system, evaluates multiple outputs, selects a final image, and does not manually alter that image afterward, who should be regarded as its author? This paper examines that question through the concept of Zero-Edit, defined here as a generative-AI artistic practice in which the selected output is deliberately preserved without post-generation manual modification. This paper does not argue that the use of artificial intelligence automatically produces human authorship, nor the reverse. It proposes instead that authorship can be evaluated through the nature, degree, traceability, and expressive relevance of human creative agency across the production process, using a five-dimension framework: conceptual, directive, iterative, evaluative/curatorial, and contextual agency, assessed alongside the AI system's expressive contribution. Draft 1.4 extends the prior version in three respects. It applies the framework to a real Zero-Edit artwork (Section 15.1–15.7), producing a differentiated result rather than a purely theoretical classification. It substantially expands the literature base underlying Sections 5–12. Draft 1.5 adds a contrasting composite case (Section 15.8) that the framework does not support — evidence that the framework can discriminate rather than only confirm.