AI Governance Ratings Methodology
Defensibility comes from how the rating is governed.
The AIGR™ Artificial Intelligence Governance Ratings methodology uses a controlled decision process: fixed scope, traceable evidence, recorded reviewer judgment, critical-gate logic, separated approval and continuing surveillance. This AI Governance Ratings methodology describes how the opinion is governed while keeping proprietary weights, thresholds and calibration logic controlled.
An analytical measure and a rating designation serve different jobs.
The analytical layer gives reviewers enough resolution to diagnose the control environment. The rating layer converts the governed assessment record into a stable executive designation without presenting the raw score as the rating itself.
AIGR Governance Score™
A 0–100 measure supporting diagnostics, remediation analysis and trend monitoring. Calculation logic is controlled.
AIGR™ AI Governance Rating
AIGR-100 through AIGR-40, plus AIGR-NR when a defensible opinion cannot be issued.
Eight domains keep the assessment broad enough for enterprise governance.
Governance & oversight
Accountability, decision rights, policy ownership and executive visibility.
Organizational readiness
Skills, operating model, training and management capability.
Risk management
Classification, impact analysis, control design and remediation discipline.
Responsible AI practices
Intended use, transparency, fairness, human oversight and recourse.
Enterprise architecture
System design, integration, data flows and technical control points.
Cybersecurity governance
Access, model and data security, supplier controls and incident readiness.
Regulatory alignment
Mapping controls to applicable obligations and jurisdictional requirements.
Operational governance
Monitoring, change management, lifecycle control and evidence continuity.
A governed decision process, not a one-time questionnaire.
Six disciplines hold the opinion together.
Scope
The subject, jurisdiction, lifecycle and evidence period are fixed before conclusions are drawn.
Evidence
Material conclusions are anchored to relevant, current, traceable and sufficient artifacts.
Judgment
Reviewer roles, exceptions and material decisions are recorded rather than implied.
Issuance
Final approval is separated from routine evidence collection and commercial pressure.
Monitoring
Incidents, material changes and evidence expiry can trigger a new review.
Critical gates
Severe control failures cannot be offset by strength elsewhere in the assessment.
Seven designations, plus Not Rated.
| Designation | Meaning | Interpretation |
|---|---|---|
| AIGR-100 | Exemplary governance | Leading maturity |
| AIGR-90 | Advanced governance | High confidence |
| AIGR-80 | Established governance | Established |
| AIGR-70 | Developing governance | Improvement required |
| AIGR-60 | Emerging governance | Material remediation |
| AIGR-50 | Limited governance | High concern |
| AIGR-40 | Critical governance concern | Urgent review |
| AIGR-NR | Not Rated | Insufficient scope, evidence or conditions |
How to read the scale: the designation is categorical. It is not a percentage, probability or credit score. The AIGR Governance Score™ is a separate analytical measure.
Five parameters define exactly what the opinion covers.
| Parameter | What is fixed | Governance effect |
|---|---|---|
| Entity | Legal entity and accountable business unit | Identifies ownership of findings and remediation |
| System | Named AI system, model/version and intended use | Prevents transfer of the rating to another system |
| Lifecycle | Development, validation, deployment or production state | Sets the evidence standard for the stage assessed |
| Jurisdiction | Applicable regulatory and operating context | Determines which obligations are mapped |
| Evidence period | Evidence window and effective date | Makes the opinion point-in-time and reproducible |
Four tests decide whether an artifact can support a conclusion.
Relevance
The artifact addresses the control requirement being tested.
Currency
The artifact reflects the current system and assessment period.
Traceability
The artifact has a verifiable source, accountable owner, date and version.
Sufficiency
The artifact shows the control operating, not merely designed or intended.
Evidence supports the requirement.
Evidence is present, but reviewer determination remains open.
Excluded only with a documented scope rationale.
A material failure capable of holding the governed outcome.
Assessment, remediation support and rating approval are treated as separate functions.
Conflicts are recorded at intake. The evidence record and reviewer challenge remain visible through the decision process so final approval can be evaluated independently of routine assessment work.
Collect and test evidence against the framework; do not determine the final designation.
Support corrective action without controlling rating approval.
Challenge the assessment record and approve, condition, decline or return the opinion.
Affirmation
Current evidence continues to support the rating.
Upgrade / downgrade
Material changes in governance capability, evidence quality or open conditions move the designation.
Conditional
An open critical gate or material condition constrains the decision state.
Suspension
An incident, material change or evidence-integrity challenge requires investigation.
Withdrawal
The rating can no longer be supported within the original scope or evidence access ends.
Not Rated
AIGR-NR applies where scope, evidence, independence or assessment conditions are insufficient.
A governed opinion needs a documented route for challenge.
Appeal is intended to correct factual or methodological error, not to negotiate the commercial consequences of a designation.
Grounds for challenge
Factual error, material evidence not considered, misapplication of the methodology, or a misstated assessment scope.
Independent review
The challenge should be considered by reviewers who were not responsible for the original decision.
Recorded outcome
The determination and rationale remain part of the rating record whether or not the designation changes.
The framework itself is version-controlled.
Versioning
Every rating identifies the methodology version used.
Change control
Material changes to criteria, gates or scale carry an effective date and rationale.
Calibration
Reviewer decisions are tested for consistency across comparable evidence.
Confidentiality
Underlying client evidence remains segregated from the public rating record.
Cohorts
Any benchmark cohort is constructed under consent and confidentiality rules.
Records
Assessment, challenge and approval records are retained to support reproducibility.
What an AIGR™ rating is—and what it is not.
| A rating is | A rating is not |
|---|---|
| A governance opinion supported by a defined evidence record | A certification, accreditation or attestation of compliance |
| Scoped to a named entity, system, period and jurisdiction | Automatically transferable to other systems or periods |
| A point-in-time signal subject to defined rating actions | A permanent or guaranteed status |
| Decision support for boards, buyers and risk teams | An audit opinion, legal opinion or regulatory approval |
| A categorical designation from AIGR-100 to AIGR-40 | A percentage, probability, credit rating or safety guarantee |
Published framework · 2026
Artificial Intelligence Governance Ratings framework.
Download the published ratings framework for the category thesis, rating architecture, lifecycle and scale. Proprietary formulas, weights, thresholds and calibration logic remain controlled.
A rating is an opinion within a defined scope, methodology version and date.
It is not a certification, accreditation, audit opinion, legal opinion, regulatory approval, compliance attestation, probability, credit rating, investment recommendation or safety guarantee.