The Five Architectural Pillars of Execution Governance
Minimum structural requirements for governable execution authority in adaptive systems.
Pillar I — Externalized Authority
Execution authority must originate outside the system being governed. Authority inferred from model behavior or optimization is indistinguishable from self-permission.
Pillar II — Pre-Execution Admissibility
All actions must pass admissibility checks before execution and before state change occurs. Governance that begins after execution is documentation — not control.
Pillar III — Runtime Custody & Stop-Rights
Responsibility without the power to halt execution at decision-time is non-binding. Stop-rights must be structurally enforced at runtime.
Pillar IV — Independent Validation
Validation must rely on signals or proofs the executing system cannot fabricate or tamper with. A system validating itself is marking its own homework.
Pillar V — Impossibility, Not Discouragement
Governance exists only when certain actions are structurally impossible. Policies, warnings, or “should not” constraints do not constitute enforcement.
Structural Principle
If any one pillar is absent, governance does not exist.