Why a doctrine — not a framework
Frameworks recommend. They describe best practices, suggest structures, and invite adoption. They are useful and often valuable.
The AI Non-Delegation Doctrine does something different. It is a normative governance instrument: it establishes what may and may not happen, defines the conditions under which authority is legitimate, and is designed to be enforced — not merely referenced.
For consultancies advising clients on AI governance, this distinction is operationally significant. A client whose governance is built on frameworks can always explain. A client whose governance satisfies the doctrine can also refuse — and prove it.
A framework says
"Here are the best practices for human oversight in AI-assisted decisions. We recommend a meaningful review process, clear accountability structures, and documentation of decision rationale."
The doctrine says
"Authority over consequential execution is not delegable to an AI system. Governance must be resolved at the commit boundary by legitimate human authority. Where this cannot be demonstrated, the system is ungoverned at that point."
What this means for your clients
The pattern emerging across jurisdictions is consistent: systems that claim human oversight but cannot demonstrate it at the point of consequence are increasingly exposed — to litigation, regulatory action, and reputational harm. MiDAS in Michigan. Toeslagen in the Netherlands. The WA seatbelt programme in Australia.
The doctrine provides your clients with a structure that is testable before something goes wrong, not only explainable after. The anti-theatre tests built into the doctrine — perimeter closure, fail-closed semantics, protected intervention — are pre-deployment questions, not incident-response tools.
"Governance that can only explain after the fact is not governance. The doctrine asks one question: at the moment this decision took effect, who held authority — and could they have changed the outcome?"
The State Zero layer
The State Zero frameworks, attributed to Toni Scorsese, Ph.D., address the human and institutional conditions upstream of any decision. For consultancies, this is the layer that most governance assessments miss: not whether a human was assigned, but whether that human was capable of independent judgment under the real conditions of the system.
Cognitive load, decision fatigue, volume pressure, anchoring to prior outputs, inherited drift, and the suppression of challenge — these are not personal weaknesses. They are governance conditions. The doctrine and its companion frameworks give you the language to assess and address them systematically.
How to proceed
Warrentor is in its pilot phase. We are not yet publicly offering consultancy-facing services, but we are in conversation with a small number of organisations whose work aligns with the doctrine's purpose.
If you are working with clients on AI governance and believe the doctrine is relevant, complete the form below. We will respond directly. There is no pitch on our end — the conversation is the first step.