The rapid rise of generative AI-generated image-based technologies is profoundly challenging the traditional copyright law paradigm, which is centered on human intellectual creation. In this technological architecture, the user's "prompt" becomes the key medium connecting human intent and machine-generated results, and its legal attributes directly relate to the copyright ability of the final generated work. This article aims to analyze existing scholarly debates, focusing on the core legal issues of prompts as a carrier of originality. Through a review of relevant theories and key judicial cases, this article clarifies that when prompts possess a certain information density, demonstrate the user’s continuous control over the generation process, and to a certain extent reflect the user’s conception of and control over the work’s expression, they can serve as important evidence of originality for the copyright of AI-generated image works.

As the development of generative Artificial Intelligence Technology continues to advance, problems related to Copyright Infringement caused by AI-generated Works are rising more frequently; In particular, User Liability due to Use of such Technologies is also On-the-rise. Systematically analyse the liability determination pathways of direct infringement, contributory infringement and fair use defences triggered by artificial intelligence-generated content in relation to substantial similarity under the specific circumstances of AI. Further demonstrate the difficulties of adapting to traditional safe harbours for AI governance. According to this, the paper presents a phased-separated and subject-based liability-sharing mechanism. Considering both types of criteria that belong to the "controls" and "Contextualisation", so as to find an appropriate way between technological innovation and copyright protection. This offers viable theoretical reference for optimising copyright systems in the AI era.