RETHINKING ONLINE INTERMEDIARY LIABILITY IN THE AGE OF GENERATIVE AI: FROM "SAFE HARBOR" TO A MODEL OF ALGORITHMIC ACCOUNTABILITY

Abstract

The purpose of the article is to prove the conceptual and functional inadequacy of the "safe harbor" doctrine for regulating the liability of generative artificial intelligence (AI) providers for copyright infringement and to substantiate the need for developing a special (sui generis) liability model. Methods. The research is based on a combination of general scientific and special legal methods. Doctrinal analysis was used to study the legal norms governing the liability of online intermediaries in the US (MCA) and the EU (E-commerce Directive, DSA, CDSM Directive). The comparative legal method was used to contrast the American and European approaches to intermediary immunity. The system-structural method allowed for the deconstruction of the generative AI content creation process into stages (data curation, model design, generation) and to define the provider's active role in each. The formal-logical method was applied to identify contradictions between the premises of the "safe harbor" doctrine and the technological reality of generative AI. Results. It has been established that generative AI providers play an active, defining, and co-creative role in the content creation process, which fundamentally distinguishes them from the passive intermediaries for whose protection the "safe harbor" regimes were created. It is demonstrated that the "notice-and-takedown" mechanism is ineffective for the dynamic nature of AI-generated content. A hypothesis is proposed about the need to replace the binary "liable/exempt from liability" framework with a sui generis model based on the principles of algorithmic accountability and tiered duties of care. Conclusions. It is concluded that the direct application of existing "safe harbor" doctrines to generative AI is a categorical error. The proposed new liability model shifts the focus from reactive response to proactive risk management, imposing a duty on developers to implement technical and organizational measures to prevent infringements. For Ukraine, in the context of European integration, this creates an opportunity to "leapfrog" over outdated stages of regulation and immediately implement an advanced model. Prospects for further research lie in developing specific technical standards for implementing tiered duties of care and in analyzing the economic consequences of introducing the proposed liability model.