The Default to Distrust diagnostic tests your board's AI governance across six dimensions, from director capability to drift and narrative risk. Your report places you on a five-stage maturity scale, names your weakest themes, sets out the first action for each, and cites the research on why each gap is a live exposure.
Headline diagnostic: 18 questions, about 14 minutes
The eighteen anchor questions, three per dimension. It is the instrument's own first step, and the quickest read of where your board stands.
Full diagnostic: 108 questions, about 81 minutes
All 108 questions across eighteen themes, one theme at a time. Each anchor is tested from the angles a director, an auditor, or a regulator would take. Save and finish later at any point.
Six dimensions
Board oversight and director capability: Visibility of AI in production, Agenda, ownership, and reporting, Director capability and challenge
Data and model provenance: Retrieval and traceability, Logging and lineage, Assurance and supply chain
Verification architecture: Evidence at the moment of output, Challenge and review, Adversarial testing and process-based verification
Autonomy boundaries: Declared limits, Change control, Cascade and containment
Human accountability: Pause and stop authority, Decision ownership, Review and capability
Drift, recovery, and resilience: Drift monitoring, Recovery and tolerance, Narrative detection and response
How it works
Choose your pass
The headline pass asks the eighteen anchor questions. The full pass adds the ninety supporting questions that test whether each anchor answer survives scrutiny.
Score what can be shown
Each statement scores from 0 to 4, from not in place to documented, tested, and improving. Where two scores are defensible, record the lower.
Read your report
Your report places you on the five-stage scale, flags where the headline answers outrun the detail, and sets out the first action for your three weakest themes.
Take it to the board
Ask each director to score independently. The spread between directors is often more informative than the average.
The scale
0, Not in place: This does not happen, or nobody can say whether it does. Nothing could be shown.
1, Partial: It happens in part or informally, and the gaps are known. Someone can describe what exists and what is missing.
2, Documented: It is written down and approved, but not yet tested or exercised. An approved policy, register, procedure, or board paper.
3, Tested: It is documented and has been tested or exercised in the last twelve months. A dated test, drill, audit, or real occurrence, with the result recorded.
4, Improving: It is tested, and at least one change has been made because of what a test or real use showed. The change, documented, with the test or event that prompted it.
The five stages
Unaware (0 to 14): AI is absent from the board agenda. Decisions are being made by systems the directors do not know are in production.
Reactive (15 to 28): The board learns about AI through incidents. Governance is after the fact, led by the general counsel or the chief financial officer rather than a named, technology-fluent director.
Compliant (29 to 44): A policy exists, mapped to the emerging regulatory minimum. Reporting is quarterly and honest. It has not been tested.
Resilient (45 to 58): Governance is integrated with risk management. Drills are run, drift is monitored, director capability is deliberately developed. The board can answer all six director questions within a week.
Architectural (59 to 72): Trust architecture is a declared strategic asset, reported to the market, underwritten by measurable controls, and understood by institutional investors as part of the thesis.
Your answers save to a private link as you go, so you can finish later, and your report goes to the email you sign up with. Contrlr never publishes an organisation's score, and benchmarks only ever appear as averages.
Built on Default to Distrust, the board-level framework for trust architecture and narrative risk by Tony Surtees AM and Jeremy Liddle.
Advising boards? Take your clients through it.
Consultants and advisers can set up a client, invite each director to score independently, run a pass with the client in the room, and receive a copy of every report and a board comparison.