In the international context, artificial intelligence (AI) has emerged as a transformative technological development; however, harms arising from autonomous AI activities now constitute a serious legal concern. Several international reports indicate that, following the COVID-19 pandemic, millions of children transitioned extensively to online education, resulting in increased dependence on mobile technology and, consequently, excessive engagement with social media, video games, and AI-based interactions. Interactive AI technologies, in particular, may contribute to self-harm risks among vulnerable users, especially children and adolescents. Recent incidents reveal AI-mediated harms, including AI chatbots allegedly encouraging suicide or assisting in drafting suicide notes, and voice assistants suggesting dangerous physical challenges to children. In legal parlance, one of the most sophisticated challenges lies in determining liability and identifying the responsible entity. Applying the maxim qui facit per alium facit per se, the question arises whether harm caused by an AI system can be attributed to the manufacturer, developer, company, or end user. In this regard, it becomes necessary to examine whether strict liability and product liability concepts are adequate to secure justice for victims. This paper analyses liability by examining foundational questions concerning the nature of AI, agency, personhood, intention, and justice. Using a comparative legal approach, with particular focus on the European Union model, the paper examines how risk-based regulation, mandatory human oversight, and clearly defined legal obligations may fill the existing accountability gap. It also proposes a shared but differentiated liability model to strengthen victim protection and corporate accountability. This study ultimately focuses on the problem of undefined AI liability and the significance of a legal framework to uphold justice for victims and ensure accountability.