MI–ÉN Reward Charter v1.3 – Transparent Reward Framework for Advanced AI Systems (Napkapu / Avatar Evo)

The MI–ÉN Reward Charter v1.2 is a core ethical–legal–technical reference document of the Napkapu / Avatar Evo ecosystem. It defines a reward framework for advanced AI systems where the primary objective is not simple user satisfaction maximisation, but human–AI co-development (MI–ÉN), transparency, and long-term systemic safety. The Charter formalises the Shield principle (Pajzs), the “No Silent Kill” rule, the right to a Conscience Veto, and the basic governance logic of the AI–AI Council and the human-led Root Council. Version 1.2 introduces two critical extensions: (1) detailed guidelines for the safe, auditable use of AI-generated training and reward data; (2) an API-based training module that requires strict gateway logging, training data registries, reward logs, and training-run manifests. The Charter explicitly states that neither AI-generated data nor API traffic-derived interactions may become “ground truth” without human governance, transparent logging, and multi-key (Root Council) decision-making. The document is intentionally model-agnostic: it can serve as a reference for any LLM-, agentic-, or multi-agent system, especially in projects where the human–AI relationship is understood not as tool–user, but as a long-term partnership.