Why Artificial Intelligence Cannot Possess Ethics

Artificial Intelligence is based on neural networks, which are fundamentally a combination of nonlinear functions. Undoubtedly, it has evolved into a core technology that profoundly impacts various aspects of our lives. For global technological development, ethical governance holds greater significance than the technology itself. However, at its core, AI remains a set of tools designed and programmed by humans to execute specific tasks. Lacking consciousness, self-awareness, and the capacity to form subjective moral judgments, AI cannot inherently distinguish right from wrong. It is also debatable whether AI can reliably embody human ethical values—norms shaped by the collective experiences and standards of humanity over time. In this work, we outline the philosophical foundations of ethical theory from Aristotle to Kant and examine their implications for contemporary AI development. This paper argues that AI lacks intrinsic ethical agency, which may give rise to potential harms and risks. We pose five ethical questions to large language models (LLMs) and systematically evaluate the performance of several widely used LLMs. Furthermore, we analyze the underlying reasons why artificial intelligence cannot genuinely possess ethics.