Glossary

Specification Reference

This glossary defines terms used across the ARAF standard specification, the Six Dimensions assessment model, the Governance Benchmark Index, certification materials, and the Decision Supply Chain framework. Definitions are stated consistently with usage in Trust Architecture: Making Autonomous Systems Bankable by Carly Martin (forthcoming). Where a term carries a specific technical meaning within the ARAF framework, that meaning governs over any general usage.

The framework is licensed under CC BY 4.0.

GBI score orientation: A lower GBI score represents stronger Governance Posture. ARAF Certified <= 1.75. ARAF Compliant <= 2.50. Remediation required > 2.50.

Canonical References

Terminology Discipline

Terms defined in this glossary are capitalised throughout the ARAF specification, assessment methodology, certification documents, and supporting materials. Capitalisation signals that the term carries a defined meaning within the framework. Synonyms for defined terms must not be used.

Use Decision Supply Chain, not alternatives such as decision pipeline, decision flow, or inference pipeline.

Use Governance Benchmark Index or GBI, not alternatives such as governance score or risk index.

Terminology discipline prevents term drift as the framework enters policy, insurance, and regulatory practice.

A

Accredited Assessor

An individual or organisation authorised to conduct ARAF Assessments under the Assessor Accreditation Standard. Accredited Assessors must satisfy qualification requirements (governance assessment competence, autonomous systems governance literacy, and ARAF methodology proficiency), independence requirements, and methodology adherence standards. When an Accredited Assessor issues a certification, it accepts professional responsibility for the quality of the assessment that produced it.

See also: Assessor Accreditation Standard, ARAF Assessment, ARAF Certification.

AI Governance Principles

Normative objectives referenced in AI governance frameworks, such as transparency, accountability, fairness, contestability, safety, and privacy. Principles describe what responsible AI governance should achieve. They do not, by themselves, measure whether the governance infrastructure required to achieve them exists.

See also: Six ARAF Dimensions, Agentic Risk Architecture Framework (ARAF), Governance Posture.

Accountability Displacement

The condition in which a link in the Four-Link Accountability Chain is unoccupied, or accountability is assumed to rest with a third party without documented assignment. Accountability Displacement creates exposure for the deploying organisation because undocumented accountability defaults to the entity at the centre of the Decision Supply Chain.

See also: Four-Link Accountability Chain, Decision Supply Chain, Liability Architecture.

Adaptive Stability

One of the Six ARAF Dimensions.

Adaptive Stability evaluates the governance arrangements that manage system change over time, including model retraining, scope expansion, and operational drift. Systems that adapt without corresponding governance updates accumulate Governance Debt. A system that has adequate governance at the point of initial deployment may have inadequate governance six months later if its governance architecture does not evolve with the system.

See also: Six ARAF Dimensions, Governance Drift, Governance Debt, Certification Lifecycle.

Agentic Bankability

The condition in which an autonomous system can be classified, governed, insured, financed, and relied upon by the institutions that must assume responsibility for its behaviour. Agentic Bankability is the threshold at which autonomous systems cross from technical capability into institutional deployability.

Agentic Bankability is satisfied when four conditions are met: identifiability, measurability, governability, and insurability.

The term is derived from the project-finance concept of bankability: the condition under which an asset meets the governance standards required for institutional financing.

See also: Four Conditions of Agentic Bankability, AE3, Agentic Risk Architecture Framework (ARAF), Governance Benchmark Index (GBI), Institutional Reliance.

Agentic Risk Architecture Framework (ARAF)

A classification framework for assessing the Governance Posture of autonomous system deployments. ARAF assesses Governance Posture across Six Dimensions (Autonomy Gradient, Data Sensitivity Exposure, Contract Infrastructure, Liability Architecture, Commercial Leverage, and Adaptive Stability), producing a composite Governance Benchmark Index (GBI) score on a scale of 1.0 to 5.0 (lower is better) and a Dimensional Profile.

ARAF is designed to operate as the governance architecture assessment layer above process-compliance frameworks such as ISO 42001 and risk methodology frameworks such as NIST AI RMF. It translates their outputs into signals that institutional capital markets, insurers, and regulators can interpret.

See also: Governance Benchmark Index (GBI), Six ARAF Dimensions, ARAF Certification Tiers, Autonomous Intelligence Oversight Committee (AIOC).

AE3 (Autonomous Action Consequences)

The category of outcomes produced by an autonomous system’s decisions without per-step human authorisation. AE3 is the liability category that existing frameworks (professional indemnity, product liability, technology errors and omissions) were not designed to address. It encompasses consequences of decisions made at machine scale, across machine-scale decision volumes, without individual human sign-off on each decision.

AE3 exposure is not limited to financial services. It is a structural consequence of autonomous decision execution in any domain where professional judgement historically carried liability.

See also: Liability Architecture, Agentic Bankability, Four-Link Accountability Chain, Multiplier Condition.

ARAF Assessed

The entry-level ARAF certification designation. Indicates that an autonomous system’s Governance Posture has been independently evaluated by an Accredited Assessor against the ARAF framework. ARAF Assessed status does not carry a GBI score threshold; any system that has undergone independent assessment qualifies. It is the minimum that institutional audiences should require before entering governance representation discussions.

See also: ARAF Compliant, ARAF Certified, Governance Benchmark Index (GBI), Institutional Reliance.

ARAF Assessment

An independent evaluation of an autonomous system’s governance architecture conducted using the ARAF methodology. The assessment produces a Dimensional Profile, GBI score, Multiplier analysis, Evidence Quality Tier classification per dimension, and certification determination where applicable.

See also: ARAF Certified, ARAF Compliant, ARAF Assessed, Dimensional Profile, Governance Benchmark Index (GBI).

ARAF Certification

A governance signal issued by an Accredited Assessor indicating that an autonomous system’s governance architecture satisfies defined thresholds under the ARAF methodology. Certification performs three functions: information compression (converting a six-dimensional assessment into a comparable signal), accountability transfer (the assessor accepts professional responsibility for the assessment), and market creation (making Governance Posture comparable and priceable across deployments).

See also: ARAF Certification Tiers, Governance Benchmark Index (GBI), Accredited Assessor, Certification Lifecycle.

ARAF Certification Tiers

The three certification levels defined by ARAF:

ARAF Assessed: Assessment completed; no minimum score required. Entry level for institutional discussions.
ARAF Compliant (GBI <= 2.50): Minimum institutional threshold. Governance Posture supports standard insurance coverage terms, enterprise procurement, and regulatory engagement.
ARAF Certified (GBI <= 1.75): Full Agentic Bankability conditions met. Governance Posture supports classification, insurance, financing, and full Institutional Reliance.

The tiers create a governance progression that institutions can track over time.

See also: Governance Benchmark Index (GBI), Agentic Bankability, Institutional Reliance.

ARAF Evidence Infrastructure Certification

A product-level designation (not an organisational assessment) available to governance infrastructure platforms (data governance platforms, AI gateways, action admission systems) demonstrating that their audit outputs satisfy ARAF evidentiary standards for authenticity, integrity, traceability, and exportability.

Organisations deploying certified Evidence Infrastructure inherit an evidentiary advantage in their own ARAF Assessments: the evidence their infrastructure produces qualifies as Tier 1 (infrastructure-generated) evidence.

See also: Evidence Infrastructure, Evidence Quality Tiers, Evidence Standard.

Assessor Accreditation Standard

The ARAF specification defining qualification, independence, methodology adherence, and accountability requirements for Accredited Assessors. The accreditation architecture was designed with specific reference to the failure mode documented in the FTC’s 2014 action against TRUSTe, where inadequate assessor oversight produced certifications that did not reflect actual practice.

See also: Accredited Assessor, ARAF Certification, Standard Governance.

Autonomous Intelligence Oversight Committee (AIOC)

The governance body responsible for autonomous system oversight within a deploying organisation. The AIOC is the organisational mechanism through which operational accountability is discharged in the Four-Link Accountability Chain. At scale, it operates as a standing committee with defined membership, meeting cadence, and decision log. At Seed stage, it may be a named founder or senior team member with documented responsibilities.

See also: Four-Link Accountability Chain, Board Risk Architecture, Reconstructability, Evidence Continuity.

Autonomous System

A software system capable of producing consequential decisions or actions without real-time human authorship for each decision. Examples include automated credit decision engines, AI agents executing operational tasks, algorithmic trading systems, and automated procurement systems.

See also: Autonomy Gradient, AE3, Decision Supply Chain, Agentic Bankability.

Autonomy Gradient

One of the Six ARAF Dimensions.

The spectrum of operational autonomy ranging from systems that execute precisely defined instructions (execution end) to systems that exercise discretionary judgement within broad parameters and adapt their own decision logic (adaptive autonomy end). The Autonomy Gradient is the foundational dimension: a system’s position on the gradient determines the governance architecture required to manage its AE3 exposure and establishes the governance burden that all other dimensions must address.

Assessed sub-factors include operational autonomy level, commitment authority, exception handling, scope boundary enforcement, and human oversight adequacy.

See also: Six ARAF Dimensions, AE3, Governance Posture, Multiplier Condition.

B

Bekier Accountability Test

The governing accountability standard applied throughout the ARAF framework, derived from ASIC v Bekier [2026] FCA 196. The test requires that three questions be answerable from contemporaneous records:

Who was responsible?
What did they know?
What did they do?

The Bekier test is the operational definition of Reconstructability. A governance architecture that cannot answer these three questions from contemporaneous evidence has failed the Reconstructability requirement.

See also: Reconstructability, Four-Link Accountability Chain, Evidence Continuity, Section 180(1) Duty of Care.

Bilateral Governance Principle

The principle that the Reconstructability obligation in autonomous system deployments applies to both sides of an institutional relationship. The deploying organisation must demonstrate that its decision-making governance is reconstructable. The institutional counterparty (insurer, financier, regulator) must demonstrate that its assessment of that governance was itself conducted through a reconstructable process.

In insurance contexts: the insured organisation needs governance architecture so the insurer can reliably assess risk. The insurer needs governance architecture so the underwriting process itself is defensible.

See also: Reconstructability, Evidence Continuity, Institutional Reliance, Data Sensitivity Exposure.

Board Risk Architecture

The formal governance framework through which a board discharges its autonomous system governance obligations. It comprises four components:

Governance mandate: board resolution establishing governance expectations for autonomous systems.
Board information standard: specification of what governance information the board receives, in what form, and at what frequency.
Independent assurance line: mechanism through which the board receives governance information not filtered through management.
Escalation protocol: mechanism through which governance matters are brought to the board between formal reporting cycles.

See also: Five Governance Questions, Section 180(1) Duty of Care, Governance Benchmark Index (GBI), Autonomous Intelligence Oversight Committee (AIOC).

C

Certification Lifecycle

The process through which an autonomous system is assessed, certified, monitored, and periodically reassessed under the ARAF framework. The lifecycle follows the structural model used in aviation for airworthiness: initial certification, event-triggered reassessment, and scheduled maintenance.

Includes initial assessment, certification issuance, scheduled reassessment (maximum 12-month interval), event-triggered reassessment (material changes trigger reassessment regardless of schedule), and potential certification withdrawal.

See also: ARAF Certification Tiers, Governance Drift, Adaptive Stability, Reconstructability.

Commercial Leverage

One of the Six ARAF Dimensions.

Commercial Leverage measures the degree to which an organisation has become operationally dependent on an autonomous system. High dependency may make governance remediation commercially difficult even when governance weaknesses are identified, and may trigger Multiplier Conditions in combination with other dimensional weaknesses. This dimension explains why governance gaps, once opened, tend to remain open.

Assessed sub-factors include revenue concentration, customer relationship embedding, remediation commercial resistance, and technology lock-in.

See also: Six ARAF Dimensions, Leverage Collapse, Multiplier Condition, Governance Debt.

Contemporaneous Production Requirement

The evidentiary principle that governance records must be created at the time decisions are made, not reconstructed after the fact. Records produced in response to a claim, investigation, or audit carry significantly lower evidential weight than records generated as routine governance output.

See also: Reconstructability, Evidence Continuity, Evidence Infrastructure, Bekier Accountability Test, Evidence Quality Tiers.

Contract Infrastructure

One of the Six ARAF Dimensions.

Contract Infrastructure governs how the system’s decisions are admitted into external environments and structured into commitments. It evaluates whether the contractual framework governing customers, vendors, and system participants reflects the operational reality of autonomous decision-making. Contracts that predate autonomous deployment, or were drafted without AI-specific provisions, typically fail this dimension. Contract infrastructure gaps are often invisible until an adverse outcome occurs and parties determine loss allocation.

Assessed sub-factors (v3.1):

D3.1 — Master Services Agreement Maturity: Do customer agreements contain AI-specific provisions, scope limitations, and liability carve-outs that reflect the operational reality of autonomous decision-making?
D3.2 — Vendor Agreement Adequacy: Do agreements with AI infrastructure providers, foundation model developers, and data suppliers address governance obligations, including model and infrastructure contracts, SLAs, indemnification structure, and provider substitution rights with political designation triggers?
D3.3 — Data Processing and Liability Provisions: Are data processing agreements in place and adequate for all deployment chain entities? Do liability provisions reflect actual AE3 exposure, and is accountability contractually allocated across the Decision Supply Chain, including political force majeure provisions for government supply chain risk designation events?

See also: Six ARAF Dimensions, Infrastructure Collapse, Decision Supply Chain, Liability Architecture.

D

Data Sensitivity Exposure

One of the Six ARAF Dimensions.

Evaluates two related risk categories: the sensitivity of operational data used by the system, and the provenance and legal basis of training data. Systems with weak training data provenance documentation face backward-compounding governance exposure as deployment matures.

Assessed sub-factors include operational data sensitivity (personal, health, financial, commercially sensitive, jurisdictionally restricted data) and training data provenance across foundation model, fine-tuning, and retrieval-augmented generation layers.

See also: Six ARAF Dimensions, Bilateral Governance Principle, Reconstructability, Evidence Standard, Contemporaneous Production Requirement.

Decision Infrastructure

The full organisational and technical system through which an organisation produces consequential decisions. Decision Infrastructure encompasses both process (the Decision Supply Chain) and structure (the Distributed Decision Infrastructure of actors, systems, and data sources traversed by the chain). Governance of autonomous systems is, in institutional terms, governance of Decision Infrastructure.

See also: Decision Supply Chain, Distributed Decision Infrastructure (DDI), Autonomous System.

Decision Supply Chain

The sequence of systems, actors, and processes through which a consequential decision is produced. The chain extends from data inputs through model inference, human review, and execution infrastructure to final action or outcome.

Governance of autonomous systems requires governance of the entire Decision Supply Chain, not only the model at its centre. Each link in the chain must be governed, and evidence must be continuous across links. The chain is only as governable as its least governed link.

See also: Four-Link Accountability Chain, Evidence Continuity, Reconstructability, Distributed Decision Infrastructure (DDI), Governance Boundaries.

Delegation Gap

The gap between authority delegated to an autonomous system by its governance architecture and authority exercised in practice. The Delegation Gap opens when systems operate at higher autonomy than governance documentation contemplates, or when behaviour drifts beyond parameters on which the original governance assessment was based.

See also: Autonomy Gradient, Governance Drift, Adaptive Stability, Four-Link Accountability Chain.

Dimensional Profile

The set of six scores produced during an ARAF Assessment representing Governance Posture across the Six ARAF Dimensions. The Dimensional Profile is the primary diagnostic output of the methodology and enables targeted remediation at dimension level rather than whole-of-system redesign.

For governance action, the Dimensional Profile is more important than the composite GBI score because it shows where risk sits and whether weakness is diffuse or concentrated.

See also: Six ARAF Dimensions, Governance Benchmark Index (GBI), ARAF Assessment, Multiplier Condition.

Distributed Decision Infrastructure (DDI)

The operational reality in which consequential decisions are produced not by a single system or actor, but by a distributed network of models, data pipelines, human reviewers, and execution infrastructure operating across organisational boundaries, employment relationships, and jurisdictions.

DDI is the governance context in which the Decision Supply Chain operates. Governance designed for hierarchical human-authored decisions does not adequately govern DDI.

See also: Decision Supply Chain, Four-Link Accountability Chain, Governance Posture, Autonomous System.

E

Evidence Continuity

The requirement that governance evidence follows a decision through each link of the Decision Supply Chain. Evidence Continuity ensures the decision record can demonstrate how each participant contributed to the final outcome. A break at any link creates a Reconstructability gap.

Chain-incomplete evidence records are the most common governance gap identified in ARAF Assessments.

See also: Reconstructability, Decision Supply Chain, Contemporaneous Production Requirement, Evidence Standard.

Evidence Infrastructure

Systems that automatically generate governance records supporting ARAF evidence requirements. Examples include AI gateways, monitoring infrastructure, data lineage platforms, and action admission systems. Evidence Infrastructure converts governance obligations into operational practice by making evidence generation a routine system function.

Records produced by Evidence Infrastructure qualify as Tier 1 evidence because they are generated contemporaneously in a form that cannot be altered retroactively.

See also: Evidence Continuity, Evidence Standard, Evidence Quality Tiers, Reconstructability, ARAF Evidence Infrastructure Certification.

Evidence Quality Tiers

The three-tier classification of governance evidence based on when and how evidence was produced. Tier classification directly affects institutional confidence and may limit certification tier achievable.

Tier 1: Infrastructure-generated. Contemporaneous, tamper-evident records produced by governance infrastructure as natural operational output. Highest confidence.
Tier 2: Contemporaneous documentation. Records produced at time of governance decisions (board papers, AIOC minutes, deployment approval records). High confidence, dependent on consistency and completeness.
Tier 3: Reconstructed documentation. Records assembled in response to assessment requests. Lower confidence and potentially certification-limiting.

See also: Evidence Standard, Evidence Infrastructure, Reconstructability, Contemporaneous Production Requirement.

Evidence Standard

The ARAF specification defining categories and quality requirements of governance evidence used during assessments. Evidence is evaluated across four categories (Design Evidence, Deployment Evidence, Operational Evidence, Outcome Evidence) and must satisfy authenticity, integrity, traceability, and exportability requirements.

The standard for ARAF evidence is not what the organisation says about its governance. It is what governance infrastructure produces.

See also: Evidence Continuity, Evidence Infrastructure, Evidence Quality Tiers, Reconstructability, Contemporaneous Production Requirement.

F

Five Governance Questions

The five questions a board must be able to ask and receive adequate answers to in order to discharge duty-of-care obligations in autonomous system governance:

What autonomous systems is this organisation operating, and at what autonomy levels?
Who is accountable for governance of each system, and is that accountability documented?
What is the evidence base for governance claims management is making?
When was Governance Posture last independently assessed, and what did the assessment find?
How will the board know if Governance Posture deteriorates between formal assessments?

See also: Board Risk Architecture, Section 180(1) Duty of Care, Governance Benchmark Index (GBI), Bekier Accountability Test.

Four Conditions of Agentic Bankability

The four conditions required for Agentic Bankability:

Identifiability: the system and decisions can be described precisely enough for governance assignment.
Measurability: Governance Posture can be assessed against a defined standard.
Governability: behaviour can be governed through documented accountability structures.
Insurability: risk exposure can be underwritten by institutional counterparties.

See also: Agentic Bankability, Governance Benchmark Index (GBI), ARAF Certified, Institutional Reliance.

Four-Link Accountability Chain

The four accountability roles that must be assigned and documented for autonomous deployment:

Design accountability: responsibility for system architecture and training.
Deployment accountability: responsibility for production deployment context.
Operational accountability: responsibility for ongoing operation and governance.
Outcome accountability: responsibility for consequences of system decisions.

Each link must be occupied by a named entity or individual with documented responsibilities. Accountability Displacement occurs when a link is unoccupied or assumed without documentation.

See also: Accountability Displacement, Decision Supply Chain, Autonomous Intelligence Oversight Committee (AIOC), Bekier Accountability Test.

G

Governance Benchmark Index (GBI)

The composite governance score produced by an ARAF Assessment. GBI translates six-dimensional governance into a comparable signal that institutional stakeholders can interpret without reviewing the full assessment. Lower scores represent stronger Governance Posture. The scale mirrors credit-rating logic: 1.0 is strongest posture, 5.0 is highest risk.

GBI functions as a compressed signal of Reconstructability. A low score indicates governance architecture can be demonstrated from contemporaneous records. A high score indicates material governance gaps.

Score thresholds: ARAF Certified <= 1.75. ARAF Compliant <= 2.50. Remediation required > 2.50.

See also: Dimensional Profile, ARAF Certification Tiers, Multiplier Condition, Reconstructability.

Governance Boundaries

The three architectural boundaries where governance controls must operate across the Decision Supply Chain:

Data boundary: governs what information enters the system (maps to D2: Data Sensitivity Exposure).
Inference boundary: governs model interaction and jurisdictional accountability for inference (maps to D1: Autonomy Gradient).
Action boundary: governs consequential actions the system may take (maps to D3: Contract Infrastructure).

See also: Decision Supply Chain, Six ARAF Dimensions, Evidence Continuity.

Governance Debt

The accumulation of unresolved governance obligations when autonomous systems are deployed without adequate governance architecture. Like technical debt, Governance Debt compounds over time: the longer a system operates without adequate governance infrastructure, the more expensive remediation becomes.

See also: Governance Drift, Delegation Gap, Adaptive Stability, Governance Posture.

Governance Drift

The gradual divergence between governance architecture originally approved for an autonomous system and the system’s operational behaviour over time. Governance Drift can occur as models are retrained, deployment contexts expand, or practices change without corresponding governance updates.

See also: Adaptive Stability, Governance Debt, Delegation Gap, Certification Lifecycle.

Governance Posture

The overall adequacy of governance architecture surrounding an autonomous deployment. Governance Posture is evaluated using the Six ARAF Dimensions and expressed through the Dimensional Profile and GBI score. It is a measurable position on a continuous scale, not a binary state.

See also: Dimensional Profile, Governance Benchmark Index (GBI), Six ARAF Dimensions, Reconstructability.

Governance Signal

The structured output produced by an ARAF Assessment that communicates Governance Posture to institutional audiences. It includes GBI score, Dimensional Profile, Multiplier analysis, Evidence Quality Tier classification, and certification determination. It is designed for interpretation by boards, insurers, investors, and regulators without requiring full reassessment.

See also: Governance Benchmark Index (GBI), Dimensional Profile, ARAF Certification, Institutional Reliance.

I

Infrastructure Collapse (Multiplier)

A Multiplier Condition in the ARAF methodology. Triggered when Dimension 3 (Contract Infrastructure) >= 4 and Dimension 1 (Autonomy Gradient) >= 3. It occurs when significant autonomy combines with weak contract infrastructure, creating governance exposure that cannot be mitigated through conventional contractual allocation.

See also: Multiplier Condition, Autonomy Gradient, Contract Infrastructure, Governance Benchmark Index (GBI).

Institutional Reliance

The condition in which institutional actors can rely on an autonomous system’s governance architecture with confidence that accountability is supported. Institutional Reliance is the operational test of Agentic Bankability: a system achieves Institutional Reliance when insurers can underwrite it, boards can govern it, investors can finance it, and regulators can supervise it.

See also: Agentic Bankability, Four Conditions of Agentic Bankability, Governance Posture, ARAF Certified.

L

Leverage Collapse (Multiplier)

A Multiplier Condition in the ARAF methodology. Triggered when Dimension 5 (Commercial Leverage) >= 4 and Dimension 4 (Liability Architecture) >= 3. It occurs when high commercial dependency combines with weak Liability Architecture, making remediation commercially difficult in practice.

See also: Multiplier Condition, Commercial Leverage, Liability Architecture, Governance Benchmark Index (GBI).

Liability Architecture

One of the Six ARAF Dimensions.

Liability Architecture evaluates how responsibility for consequences of autonomous decisions is allocated between deploying organisation, customers, vendors, and other Decision Supply Chain participants. Effective Liability Architecture requires explicit treatment of AE3 exposure in contractual and governance frameworks.

Assessed sub-factors include recognition of AE3 as a distinct category, liability cap adequacy, carve-out structure, and insurance coverage for autonomous action consequences.

See also: Six ARAF Dimensions, AE3, Decision Supply Chain, Leverage Collapse, Systemic Escalation.

M

Multiplier Condition

A condition where combinations of dimensional weaknesses amplify governance exposure beyond what a simple average suggests. Three Multiplier Conditions are defined:

Systemic Escalation: D1 >= 4 and D4 >= 4.
Infrastructure Collapse: D3 >= 4 and D1 >= 3.
Leverage Collapse: D5 >= 4 and D4 >= 3.

Individual weaknesses may be manageable. Compound weaknesses create systemic exposure because compounding reduces remediation options and increases remediation cost.

Multiplier Conditions are reflected in GBI interpretation and capture non-linear governance risk.

See also: Infrastructure Collapse, Leverage Collapse, Systemic Escalation, Governance Benchmark Index (GBI), Dimensional Profile.

R

Reconstructability

The ability to demonstrate, using contemporaneous evidence, how an autonomous system produced the decision that resulted in the outcome under examination. Reconstructability is the organising principle of the ARAF evidence standard: evidence requirements, documentation obligations, and accountability architecture are designed to ensure decisions are reconstructable.

A system whose decisions can be reconstructed from contemporaneous records has demonstrable governance. A system whose decisions must be reconstructed after the fact has failed the Reconstructability requirement regardless of technical performance.

Institutions can rely on autonomous systems only to the extent decisions can be reconstructed from contemporaneous governance records.

See also: Bekier Accountability Test, Evidence Continuity, Contemporaneous Production Requirement, Bilateral Governance Principle, Evidence Standard, Evidence Quality Tiers.

S

Section 180(1) Duty of Care

The duty of care and diligence imposed on directors and officers by section 180(1) of the Corporations Act 2001 (Cth). In autonomous governance, this requires informed board oversight of material autonomous systems. Reliance on unverified management assurances may fail this duty where deployment is material in scale.

See also: Five Governance Questions, Board Risk Architecture, Bekier Accountability Test, Governance Benchmark Index (GBI).

Six ARAF Dimensions

The six structural assessment dimensions used in the Agentic Risk Architecture Framework to evaluate the governance architecture of an autonomous system deployment. Each dimension corresponds to a structural property of the Decision Supply Chain:

Autonomy Gradient: system operational autonomy level.
Data Sensitivity Exposure: sensitivity and governance of system data.
Contract Infrastructure: adequacy of contractual framework.
Liability Architecture: allocation of responsibility for consequences.
Commercial Leverage: operational dependency on the system.
Adaptive Stability: governance of system change over time.

Each dimension is scored individually to produce the Dimensional Profile. Multiplier Conditions may apply when combinations indicate amplified governance risk. Together, the dimensions produce a measurable governance signal (the GBI) that institutional actors can rely upon.

The relationship: Principles describe governance objectives. ARAF dimensions measure whether the governance architecture required to meet those objectives is present, measurable, and certifiable.

See also: Dimensional Profile, Governance Benchmark Index (GBI), Multiplier Condition, Agentic Risk Architecture Framework (ARAF).

Standard Governance

The processes through which the ARAF standard is maintained, versioned, and updated. Standard Governance covers publication and maintenance, methodology updates, interpretive guidance, and stewardship of the specification. Standard Governance does not constitute assessor accreditation, certification issuance, or statutory authority.

Governance of the standard and operation of certification functions are related but distinct, following structural separation seen in ISO and PCI DSS ecosystems.

See also: Version Discipline, Assessor Accreditation Standard, Certification Lifecycle.

Systemic Escalation (Multiplier)

A Multiplier Condition in the ARAF methodology. Triggered when Dimension 1 (Autonomy Gradient) >= 4 and Dimension 4 (Liability Architecture) >= 4. It occurs when high autonomy combines with weak Liability Architecture, creating rapidly scaling exposure.

This is the most severe compound exposure: high-speed autonomous decisions producing consequences that no liability framework is designed to absorb.

See also: Multiplier Condition, Autonomy Gradient, Liability Architecture, AE3.

T

Trust Architecture

The governance infrastructure required to make autonomous systems institutionally governable. Trust Architecture converts autonomous systems from opaque operational risk into assets that can be classified, governed, insured, financed, and relied upon. The term encompasses structural design of governance systems and institutional standards used to assess that design.

See also: Agentic Bankability, Agentic Risk Architecture Framework (ARAF), Governance Posture, Institutional Reliance.

V

Version Discipline

A governance principle requiring each standard version to be clearly identified by version number, publication date, and changelog, and maintained in stable citable form. Version Discipline ensures assessments conducted under a specific version remain interpretable against that version, and that changes are transparent to the market.

Institutional audiences citing the standard in procurement requirements, regulatory guidance, or underwriting guidelines must be able to reference a specific version with confidence that content is stable.

See also: Certification Lifecycle, Standard Governance, Agentic Risk Architecture Framework (ARAF).

Citation

Martin, Carly. Agentic Risk Architecture Framework (ARAF), Version 3.0. Institute for Autonomous Governance Pty Ltd, 2026.