<p>This article intervenes in current debates on AI ethics by introducing the concept of amorphization of moral norms—the loss of metanormative stability under algorithmic recontextualization, visibility filtering, and automated trade-offs. Unlike ordinary context-sensitivity governed by metanorms, this phenomenon refers to conditions in which agents can no longer reliably predict which metanorm governs norm selection. Norms persist but lose positional stability. A criterion that is in principle testable is proposed, operationalised through testing whether expert users with full system access can reliably forecast norm-selection patterns. The argument further holds that AI does not possess moral agency in the classical sense but functions as an infrastructural precondition for the reconfiguration of normative hierarchies—in an empirical rather than transcendental sense. Engaging with key strands of AI ethics, the analysis identifies a systematic gap in current literature: while existing accounts explain who is responsible and how agency is attributed, they do not explain how the normative framework within which these questions arise becomes structurally unstable. The central ethical challenge of AI is therefore not whether machines can become moral agents, but how algorithmic infrastructures transform the stability, hierarchy, and applicability of moral norms themselves.</p>