Governance Doctrine · Not a Framework
Every day, consequential decisions are made about people by systems that claim to be authorised, reviewed, and correct. Warrentor gives you a structured way to ask whether that claim holds.
This is not legal advice. Warrentor examines whether authority was properly exercised — it does not represent you in any legal proceeding.
The five questions that matter
Documented Cases
These cases show that the failure pattern is not accidental. The same structural collapse appears across jurisdictions, systems, and sectors.
Michigan's automated system accused 40,000 people of unemployment fraud without meaningful human review. Cases were generated, reviewed, and actioned by the same system that produced the original determination.
"The system claimed to have reviewed. What it produced was confirmation — not decision."
The Dutch Tax Authority used automated systems to flag families for benefit fraud. Authority was distributed across systems and officials in ways that made no single person responsible for the decision that devastated thousands of families.
"When responsibility is shared by everyone, it belongs to no one."
Western Australia's AI seatbelt detection system issued fines where the claimed human review — seconds per image, under volume pressure — could not constitute genuine independent judgment. A human was formally present. The question is whether they could actually exercise authority.
"Presence is not the same as capability. Authority requires both."
Live Activity
Anonymised interactions by organisation category and decision type. Named data is available to the administrator.
Top 10 interactions
This section will display anonymised interaction statistics once the Warrentor database is live. Organisation names are suppressed in the public view. Individual records are deleted after one year; aggregate counts are retained.
How the assessment works
The process is structured, not open-ended. Each stage builds on the last. You remain in control throughout.
You tell us what happened — the decision that affected you, the organisation, the outcome, and roughly when it occurred. No technical knowledge required.
We help you identify what authority was claimed — who was said to be responsible, what review was asserted, and what the organisation says happened. This is the baseline we test against.
The five questions from the doctrine are applied to your specific situation — as an examination of whether the formal claim of authority holds under scrutiny.
We identify which of the three failure modes applies — absence, degradation, or fragmentation — and what that means for the decision you received.
You receive a structured summary of the authority analysis, what questions to ask, and where relevant, what escalation options exist. We do not tell you what to do. We give you the analysis to act on.
Warrentor does not make decisions for you. Every interpretation, every escalation choice, and every action is yours. You are always the human in the loop.
"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
Foundational Architecture
A framework recommends. A doctrine constrains. The AI Non-Delegation Doctrine v2.0 is a normative governance instrument, not a set of best-practice suggestions.
The doctrine establishes that authority over consequential decisions may not be delegated to an AI system. Governance must be resolved at the point of commitment by legitimate human authority under enforceable constraints.
It is technology-agnostic, jurisdiction-neutral, and deliberately not reducible to documentation, auditability, or retrospective explanation.
State Zero-A and State Zero-B address the upstream conditions that determine whether authority is usable at the moment of consequence — not merely granted on paper.
A person may hold the right title, possess the required training, and be formally assigned to review a decision — and still be unable to exercise real judgment if the conditions have been degraded.
No duly qualified human was present at the moment of consequence. The decision was produced by the system and not reviewed by anyone capable of changing it.
A human was formally present, but the conditions — volume, time pressure, information gaps, cognitive load — made independent judgment impossible in practice.
Responsibility was distributed across people, systems, and processes in ways that ensured no single person was accountable for the decision that affected you.
Who Warrentor is for
Whether you are affected by a decision, advising someone who is, or implementing governance for an organisation — the central question is the same.
You received a decision that affected your benefits, your fine, your claim, your account. You want to understand whether that decision was legitimately made — and what to do about it.
Start the assessment Got a letter? Photo shortcut →The doctrine provides a citation-baseline for authority analysis in disputes involving automated or AI-assisted decisions. Access to doctrine, companion book, and case analysis is available on request.
Request accessImplementing the doctrine in your governance architecture. The core doctrine is publicly accessible via SSRN. Further implementation guidance is available on request.
Access the doctrineWorking with clients on AI governance, accountability architecture, or responsible deployment? The doctrine provides a normative foundation that frameworks alone cannot.
Find out morePilot Programme
Warrentor is in its pilot phase. Every assessment you complete, every incident you report, and every piece of feedback you leave directly shapes what this becomes. Your experience here is not incidental — it is the evidence base.
