생성형 AI서비스제공자의 법적 책임과 의무

Abstract

AI is efficient for humans, but as a black box, AI can't clearly explain its results. In this sense, AI is ambivalent. As an unknown technology, it is difficult to predict how it will evolve, so it is necessary to hold those who develop or provide AI services accountable and responsible. However, we should not regulate the technology, but rather the service or business model. The goal is to ensure that AI services are provided reliably and that the content is not problematic. Even if service providers monitor their services, it is difficult to do so perfectly. Therefore, the law imposes certain obligations on service providers and holds them responsible for their omissions. This is the principle of OSP liability stipulated in the Copyright Act and the Information and Communications Network Act. By entering a safe harbor, OSPs can be immunized in certain cases. The safe harbor regulation has been positive for the development of the internet. Generative AI is often criticized. It's time to consider whether it applies to AI service providers as well. According to the interpretation of the Copyright Act and the Information and Com- munication Network Act, a service provider that provides generative AI services is a service provider that provides a tool for content production. Users give instructions through prompts or utilize it as a tool to create content. To clarify, the user uses the service to create the result that the user intends or desires. The scope of liability for AI services varies depending on the legal nature of the provider. Reviewing the legal status of the AI service provider is important for the activation of AI services. AI is considered a tool, and the rights to its output should belong to the user. Logically, the user is also responsible for the creation of infringing works. Since the service is ultimately created by the user, it is difficult to hold the service provider responsible for the content of the product. However, service providers are subject to a certain duty of care in that they are the developer of the service and the provider of the generative AI service. It is desirable for the development of the AI industry to impose a duty of care that requires certain actions on the part of the service provider for defects in AI, such as the continuous creation of infringing works or the illusion of inaccurate content.