Artificial Intelligence vs. Art and Copyright: Misappropriation?

The Rise of AI-Generated Ghibli Styles Reopens the Debate on Copyright and the Trivialization of Culture

The popularization of the ChatGPT image generator, capable of recreating recognizable artistic styles such as Studio Ghibli, has sparked a new wave of debate in the legal and ethical spheres. Beyond the viral phenomenon, the implications of this technology touch on deeper issues regarding copyright, unfair competition, the trivialization of art, and cultural extractivism.

Some AI platforms are using copyrighted works to train AI models without compensating the creators, while other businesses are trying to establish compensation mechanisms. The debate revolves around whether this is a case of misappropriation or a new way to generate shared wealth. Emerging regulations in the EU and the US could shape the future of this sector. Experts from the Universitat Oberta de Catalunya (UOC) analyze, from a legal perspective, the violations arising from the widespread use of recognizable visual styles, and, from an artistic perspective, warn of cultural trivialization and the loss of artistic identity in the face of aesthetic standardization driven by AI.

Copying without copying: the legal challenge of protecting a style

The visual style of a work is not directly protected by copyright, either in Spain or in the European Union. However, when a set of aesthetic features is so identifiable that it acts as a "visual signature," it may enjoy indirect protection. "Even if it is not registered as a trademark, the law protects that identity to prevent other companies from taking advantage of its reputation and confusing the public," explains Begoña González Otero, a professor at the Faculty of Law and Political Science at the UOC.

From a legal perspective, the widespread use of Studio Ghibli's style by AI platforms opens the door to potential claims for unfair competition and appropriation of artistic identity. "Massively copying a style so closely associated with its creator or studio may be illegal, even if a specific work is not copied," González Otero points out. According to Article 6 of the Unfair Competition Law (LCD), causing confusion or exploiting another's reputation constitutes an unlawful practice. There may also be a risk of dilution of an unregistered trademark if the style is proven to be distinctive and well-known. The situation changes depending on who generates the images. "When there are platforms or companies that use recognizable styles to train or produce content that competes in the market, there is a clear risk of unfair exploitation," the expert states. However, if these are individual users creating images for private use, "any direct action against them would be disproportionate," according to Directive 2004/48/EC.

Furthermore, the passage of the Digital Services Act (DSA) introduces new obligations for certain AI platforms. Specifically, these rules apply when they act as intermediaries, that is, if they allow users to generate, share, or distribute online content, such as images, text, or videos. In these cases, the DSA—which expands the framework of the E-Commerce Directive—requires platforms to act diligently against illegal content, to be more transparent, and to better protect users' rights. If the platform exceeds 45 million users in the European Union, it is considered a very large online platform (VLOP) and must meet additional requirements, such as conducting systemic risk assessments (e.g., information manipulation or mass dissemination of disinformation).

However, if a platform does not simply host third-party content but uses that content to retrain its own models, it ceases to be a mere intermediary. In that case, it assumes direct liability, which may have stricter legal consequences.

Despite this regulatory framework, practical application remains limited. González Otero warns that, "although the legal foundations exist, enforcement in Europe is weak, and the market favors opportunistic behavior." Cases such as GEMA v. OpenAI illustrate the difficulty of defending copyright against AI models.
Source