A governable AI system spans
energy to human authority.
Seven layers, six domains each, inside an institutional field. Governance is not the eighth layer — it is the filament threaded through all of it, and it only counts where a control, an owner, an authority and evidence can be named.
Field = condition. Cake = posture. The field states what is true of the environment; the cake states what this system does about it.
Capacity constrains governance; governance constrains the system. A control that is only specified is a gap, not a control — and Lapsed is reachable from any state.
Knowledge & Intelligence Runtime
Inference is not authority.
Evaluation, Calibration & Validation
The token list records what the system touches. This records what happens when it is wrong — severity, not action type. It cannot be inferred from the tokens, and it is the axis that most determines the derived tier.
The Human–System Capability Gap
When system capability exceeds verifiable human and institutional capability, effective governance can degrade even when formal controls remain present.
AI can erode the very human capability required to govern it — through deskilling, automation dependence, attention erosion, and reduced contestability.
Control → Owner → Authority → Evidence
Accountability recorded without authority — verify this is not a moral crumple zone.
Architecture Synthesizer
Configure the layers, then run the analysis. Findings describe configuration and gaps — they are not an assurance statement.
Layers define primary architectural domains; real systems span several. RAG sits across L4–L5. An AI factory spans L1–L3. Radiology AI runs L4 through L7. Governance is different: it is not a cross-layer technology but the transverse filament operating across all of them.
Six functional slices mapping the six governance functions across the entire AI lifecycle. Click a slice to explore.
Governance
The crucial distinction: The Governance Pizza is not the RATSe™ Pizza.
There are two different sets of six:
P.R.I.M.E. → AI Profiler → RATSe 1 → RATSe 2 → HCG → Outcome Governance
Stewardship | Accountability | Transparency | Safe/Secure/Resilient | Ethical/Equitable | Environmental
The RATSe™ six pillars operate inside RATSe™ 1 and RATSe™ 2, rather than becoming the six lifecycle slices. That is the clean architecture.
Comments
Post a Comment