What the doctrine provides
The AI Non-Delegation Doctrine v2.0 establishes that authority over consequential decisions is not delegable to an AI system. Governance must be resolved at the commit boundary — the moment of consequential effect — by legitimate human authority under enforceable constraints.
The doctrine is deliberately technology-agnostic and organisation-neutral. It is not a framework offering best-practice suggestions. It is a normative instrument designed to be referenced, applied, and enforced.
Authority analysis in disputes
The doctrine provides structured language for examining whether a duly qualified human was present, capable, and able to intervene at the moment a consequential decision took effect.
Admissibility and review
Detection by a system does not constitute admissibility. Review of a system's output is not equivalent to an independent determination. The doctrine maintains this distinction throughout.
Reviewability vs. legitimacy
A decision that is formally reviewable is not necessarily legitimate. The doctrine distinguishes between the existence of an appeal pathway and the existence of genuine authority at the point of original commitment.
Appeal vs. authority
The existence of an appeal does not retroactively create authority that was absent at the moment of the original decision. The doctrine is explicit on this point and provides the structural language to argue it.
The State Zero frameworks
State Zero-A and State Zero-B — attributed to Toni Scorsese, Ph.D. (AI Non-Delegation Doctrine v2.0, attribution register) — address the upstream human and institutional conditions that determine whether authority is usable at the moment of consequence.
State Zero-B (perceptual integrity before authority is granted): A person who inherits a system already shaped by drift, shortcuts, or degraded norms may no longer recognise deviation as deviation. Authority granted to a drift-normalised inheritor is not the same as authority granted to someone who can perceive the situation clearly.
State Zero-A (protected intervention before authority is exercised): It is not sufficient for an organisation to permit override, escalation, or challenge. Intervention must be expected and protected in practice. Where intervention carries career risk, reputational penalty, or informal suppression, the formal permission is ornamental.
"A human presence does not prove human authority. Authority only matters when the person can still use it."
AI Non-Delegation Doctrine v2.0 — Frank C. Schouten — 20 March 2026
Full text: SSRN 6206859
The Companion Book
A Companion Book to the doctrine is available to legal professionals on request. It contains worked examples, applied scenarios, diagrams, and explanatory guidance. It is explicitly subordinate to the doctrine and does not add obligations or amend the core.
The Companion Book is currently in pre-release. Requests are reviewed individually. Where the case or jurisdiction warrants it, Frank C. Schouten or Toni Scorsese, Ph.D. may be available for further consultation.
The public SSRN text
The full doctrine is publicly accessible at SSRN abstract 6206859. No account or payment is required to access the doctrine text. The SSRN version is the canonical citation source.