AI Risk Beyond Ethics: A Structural Framework for Design and Stability

Current discussions on artificial intelligence are primarily framed in terms of ethics, governance, and alignment. These approaches address important concerns, but they remain focused on how AI systems are used and regulated. This work argues that such a framing is incomplete. AI risk does not arise solely at the level of application. It is also embedded in the structural properties of system design—particularly in processes of convergence, coupling, and large-scale integration. To address this gap, this paper proposes a four-part framework that distinguishes between application-level and design-level risks. It develops this distinction by: establishing the non-transferability of human responsibility in technological systems, identifying a missing layer in current AI discourse, introducing the System Instability Window (SIW) as a structural phase of large-scale systems, and outlining design principles aimed at maintaining structural stability. Taken together, the framework suggests that the central challenge of AI is not only how systems should behave, but how they are structured and that without addressing design-level conditions, systemic instability cannot be fully understood or resolved. AI governanceAI safetyAI ethicssystem designsystemic riskstructural stabilitycomplex systems