This article develops an ethical legal framework for reintroducing responsibility into executable governance. Predictive systems, by generating authority without agents, displace accountability and leave institutions without appeal mechanisms. Building on the concepts of spectral sovereignty, null subjects, and the codex of authority, the paper introduces the notion of accountability injection as a design principle. It formulates a three tier model: (1) human, where non delegable critical decisions are tied to named subjects; (2) hybrid, where human judgment co exists with model output under calibrated thresholds; and (3) syntactic supervised, where delegation is permitted only with immutable ledgers, traceability, and automatic escalation triggers. Through applied case studies in EU AI Act conformity assessment, DAO governance, predictive credit scoring, and automated medical audits, the framework demonstrates how appeal and responsibility can be restored without undermining institutional efficiency. The conclusion argues that accountability must be compiled directly into the regla compilada of governance systems, creating a normative blueprint for legislators, courts, and regulators to maintain responsibility in predictive societies. DOI Primary archive: https://doi.org/10.5281/zenodo.17106808 Secondary archive: https://doi.org/10.6084/m9.figshare.30112711 SSRN: Pending assignment (ETA: Q3 2025)