Decision Lineage Protocol (DLP)

The Decision Lineage Protocol (DLP) is an AI-native governance substrate — a formal architecture for capturing organizational decisions as state transformations with full lineage, enforced by behavioral invariants and typed by epistemic provenance. DLP is built on 19 irreducible primitives organized across five tiers and governed by 10 behavioral invariants. The architecture derives from a five-step chain grounded in established theory: Conant-Ashby's theorem that governance must be a model, Ashby's law of requisite variety, Beer's Viable System Model, LeCun's world model architecture, and a formal proof that log-loss optimization breaks the symmetries governance must preserve. The protocol is independently implementable. Any organization can read this specification, build a conformant implementation, and operate governance without licensing any product from any vendor. Read the specification online: decisionlineageprotocol.io