At the core of the proposed framework is a complete prohibition on AI-driven adjudication