This concept paper addresses structural limits of control-based AI alignment approaches as systems increase in autonomy and internal complexity. It proposes resonance-based alignment as a complementary paradigm, focusing on inner coherence rather than external compliance. The paper introduces an explicit differentiation between model structure, latent dynamics, and actual effects, and argues that this distinction enables governance frameworks to better anticipate long-term risks, trustworthiness, and societal impact. The approach is presented as governance-relevant and meta-ethical, without prescribing a specific cultural value system.