Where Is The Boundary? : Comparing Intellectual Property Governance Across DALL-E, Midjourney & Stable Diffusion

Sammanfattning

The rapid development of generative AI image platforms has created new possibilities for visual content production, while also raising legal and ethical concerns related to intellectual property (IP). In particular, debates have emerged around artistic style imitation and the generation of copyrighted characters. While public discussion has largely focused on lawsuits, artist protests, and company policies, less is known about how these platforms respond to such prompts in practice. This study examines how selected generative AI image platforms operationalise IP boundaries through their responses to user prompts. The methodology consists of three phases: (i) case study materials containing legal disputes, policy texts, company statements, and documented cases to identify key tensions; (ii) the development and application of a structured prompt-testing protocol as a methodological tool for examining artistic style imitation and copyrighted characters; and (iii) a comparative analysis of platform responses. The findings show that DALL·E applies the most visible safeguards by blocking or redirecting explicit references to artists and copyrighted characters. Midjourney generally allows similar prompts while placing responsibility on users, whereas Stable Diffusion, especially through third-party access, shows limited visible restriction. These differences indicate that IP boundaries are not applied uniformly, but emerge through platform-specific forms of governance embedded in system responses and design choices. This study contributes to research on generative AI and IP by shifting attention from legal and public debate to platform behaviour in use. It also introduces a structured prompt-testing protocol that supports systematic analysis of how generative AI image platforms regulate creative requests in practice.