Cabier Consulting
    CABIER
    TRUST.INFRASTRUCTURE.FOR.REGULATED.MARKETS
    Platform
    AI Assurance
    Solutions
    Intelligence
    Academy
    Company
    Sign InGet Started
    Module · AI Controls Library

    The market has AI policies. It does not have AI controls.

    A policy states an intention. A control has an objective, a statement, a test procedure, an evidence artefact and a jurisdiction it answers to. This library is written once and mapped many times — the same control satisfies the EU AI Act, SR 11-7 and JFSA, with the delta each supervisor adds recorded rather than rediscovered.

    320
    AI controls
    17
    Control domains
    10
    Jurisdiction sets
    3
    Test methods

    Domain

    Jurisdiction

    AIG-01

    A named executive is accountable for every AI system in production.

    Each AI system carries a single named accountable executive, recorded in the inventory, with the accountability reviewed whenever ownership or materiality changes.

    Method
    Examine · Interview

    Evidence
    AI system inventory extract showing accountable owner, with board or committee minute approving the allocation.

    Jurisdiction delta

    • European Union — Article 26 deployer obligations require assigned human oversight with competence and authority evidenced.
    • United Kingdom — SMCR requires the accountability be traceable to a Senior Manager Function with a Statement of Responsibilities.
    • United States (prudential) — SR 11-7 requires model ownership distinct from independent validation ownership.
    • Canada — E-23 requires the accountable party be documented in the enterprise model risk framework.
    • Japan — JFSA expects accountability documented in Japanese for the parent board, even where the system is deployed offshore.
    • International standards — ISO/IEC 42001 Clause 5.3: roles, responsibilities and authorities within the AI management system.

    AI Governance & Accountability

    AIG-04

    Every AI system is risk-classified per jurisdiction, not once globally.

    AI systems are classified against each applicable jurisdictional risk taxonomy, and the highest applicable classification drives the control set applied.

    Method
    Examine · Test

    Evidence
    Per-system classification record with jurisdiction, classification, rationale and reviewer.

    Jurisdiction delta

    • European Union — Annex III high-risk determination with prohibited-practice screen under Article 5.
    • United States (markets) — No statutory taxonomy. Classification is driven by investor-impact and supervisory-examination exposure.
    • Singapore — MAS FEAT materiality assessment across fairness, ethics, accountability and transparency.
    • Canada — E-23 risk rating drives validation depth and frequency.
    • Gulf sovereign — National AI charter classification where the system touches citizen services or sovereign data.

    AI Governance & Accountability

    AIG-09

    Human oversight is a recorded point in the decision path, not a stated intention.

    For every material AI-influenced decision, the system records who could override, whether an override was available at the time, and whether one occurred.

    Method
    Examine · Test

    Evidence
    Decision log sample showing oversight point, reviewer identity and override disposition.

    Jurisdiction delta

    • European Union — Article 14 requires oversight measures be built into the system design, not applied procedurally afterwards.
    • United Kingdom — FCA Consumer Duty requires oversight adequate to evidence good customer outcomes.
    • United States (prudential) — Effective challenge must be demonstrable, with evidence the challenger could have changed the outcome.
    • Japan — METI guidelines expect oversight capability proportionate to societal impact.

    AI Governance & Accountability

    MDL-02

    No model reaches production without independent validation.

    Validation is performed by a party independent of development, covers conceptual soundness, outcomes analysis and ongoing monitoring design, and is signed before deployment.

    Method
    Examine · Interview · Test

    Evidence
    Validation report with independence attestation and pre-deployment sign-off date preceding the release record.

    Jurisdiction delta

    • United States (prudential) — SR 11-7 core requirement; validation scope and independence are examined directly.
    • Canada — E-23 requires validation depth to scale with the assigned risk rating.
    • European Union — Article 17 quality management system evidences equivalent pre-market conformity.
    • United Kingdom — SS1/23 requires the validation function have sufficient standing to challenge.
    • International standards — ISO/IEC 42001 Clause 8 operational planning and control.

    Model Lifecycle

    MDL-07

    Drift is detected before it becomes a supervisory finding.

    Performance, population and concept drift are monitored on a defined cadence against pre-set thresholds, with breach triggering revalidation rather than a note.

    Method
    Examine · Test

    Evidence
    Drift monitoring output with threshold definitions and a worked example of a breach-triggered revalidation.

    Jurisdiction delta

    • United States (prudential) — Ongoing monitoring is an explicit SR 11-7 pillar alongside validation.
    • European Union — Article 72 post-market monitoring plan for high-risk systems.
    • Singapore — MAS expects monitoring proportionate to the FEAT materiality assessment.
    • Canada — E-23 requires monitoring results feed the model risk report to senior management.

    Model Lifecycle

    MDL-11

    Retirement is as controlled as deployment.

    Model decommissioning follows a defined path covering dependency discovery, evidence retention, downstream notification and inventory closure.

    Method
    Examine

    Evidence
    Decommission record with dependency map and retention schedule reference.

    Jurisdiction delta

    • European Union — Article 18 requires documentation retained for ten years after placing on the market.
    • United States (markets) — Books-and-records retention applies where the model informed a recommendation.
    • Japan — APPI retention and deletion obligations apply to the training and inference data held.

    Model Lifecycle

    DAT-03

    Training and inference data lineage is reconstructable.

    Each dataset used for training, fine-tuning or retrieval carries a provenance record covering source, licence, consent basis, transformation history and residency.

    Method
    Examine · Test

    Evidence
    Dataset provenance record traced end to end for a sampled model version.

    Jurisdiction delta

    • European Union — Article 10 data governance for high-risk systems; GDPR Article 6 lawful basis alongside.
    • Japan — APPI requires purpose-of-use disclosure and consent handling for personal data at source.
    • New York State — Part 500 requires the data inventory extend to non-public information used in AI systems.
    • Gulf sovereign — Residency constraints may prohibit training data leaving the jurisdiction entirely.
    • International standards — ISO/IEC 23894 data quality and representativeness guidance.

    Data & Provenance

    DAT-08

    Cross-border AI processing is mapped and lawful.

    Every cross-border transfer arising from model training, hosting or inference is mapped to a transfer mechanism, with the mechanism re-tested when the hosting region changes.

    Method
    Examine · Interview

    Evidence
    Transfer register with mechanism, region and last review date per data flow.

    Jurisdiction delta

    • European Union — Chapter V transfer mechanism required; model hosting region is a transfer trigger.
    • Japan — APPI Article 28 third-country transfer disclosure to the data subject.
    • Canada — OSFI B-10 expects the transfer be reflected in the third-party risk assessment.
    • Gulf sovereign — Sovereign residency may require in-country inference with no egress.

    Data & Provenance

    AGT-01

    Agent authority is bounded and enumerable.

    Each autonomous or semi-autonomous agent has a declared authority envelope listing the systems it may reach, the actions it may take and the value or volume ceilings that stop it.

    Method
    Examine · Test

    Evidence
    Agent registry entry with authority envelope and a test showing an out-of-envelope action refused.

    Jurisdiction delta

    • European Union — Article 14 oversight must remain effective where the system operates autonomously.
    • United States (markets) — Supervisory obligations under FINRA 3110 extend to activity initiated by an agent.
    • United Kingdom — SMCR accountability is not discharged by delegating the action to an agent.
    • United States (prudential) — Agent actions constitute model output for SR 11-7 purposes where they influence decisions.

    Agentic Systems

    AGT-05

    Every agent action is attributable and reversible where material.

    Agent actions write an immutable log entry carrying the initiating principal, the authority basis, the tool invoked and the reversal path where the action is materially consequential.

    Method
    Examine · Test

    Evidence
    Immutable log sample with reversal executed against a test transaction.

    Jurisdiction delta

    • United States (markets) — Books-and-records treatment applies to agent-initiated communications and orders.
    • European Union — Article 12 automatic logging over the lifetime of the high-risk system.
    • Singapore — MAS expects auditability sufficient for a supervisor to reconstruct the decision.

    Agentic Systems

    AGT-09

    Agent-to-agent chains do not launder accountability.

    Where one agent invokes another, the accountability chain is preserved end to end and the weakest authority in the chain governs the action.

    Method
    Examine · Test

    Evidence
    Chain trace for a multi-agent workflow showing the effective authority applied.

    Jurisdiction delta

    • European Union — Deployer remains accountable regardless of the number of intermediating systems.
    • United Kingdom — Senior Manager accountability persists across delegated automated chains.
    • United States (prudential) — Chained models are in scope of the model inventory individually and as a system.

    Agentic Systems

    GEN-02

    System prompts are governed artefacts.

    System prompts, guardrail instructions and retrieval configurations are version-controlled, change-approved and tested as controls in their own right.

    Method
    Examine · Test

    Evidence
    Prompt version history with approval record and regression test results per version.

    Jurisdiction delta

    • European Union — Forms part of the Article 17 quality management system for high-risk systems.
    • United States (prudential) — Prompt changes constitute model change and trigger the change-management path.
    • Singapore — FEAT transparency expects the behavioural constraints be documented.

    GenAI & Prompt Controls

    GEN-06

    Output is constrained where it could constitute advice or a representation.

    Generated output that could be read as regulated advice, a price, a term or a commitment passes a constraint layer with a recorded disposition before reaching a customer.

    Method
    Examine · Test

    Evidence
    Constraint layer configuration plus sampled outputs with dispositions.

    Jurisdiction delta

    • United States (markets) — Communications with the public rules apply to generated content unchanged.
    • United Kingdom — Consumer Duty and financial promotion rules apply to the output, not the technology.
    • European Union — Article 50 transparency where a person interacts with an AI system.
    • Japan — JFSA expects suitability obligations survive automation of the interaction.

    GenAI & Prompt Controls

    CYB-21

    Adversarial exposure is tested, not assumed.

    Prompt injection, jailbreak, data-poisoning and model-extraction exposure are tested on a defined cadence and after material change, with findings tracked to closure.

    Method
    Test · Examine

    Evidence
    Red-team report with findings, severity, owner and closure evidence.

    Jurisdiction delta

    • New York State — Part 500 penetration testing and vulnerability management extend to AI-enabled systems.
    • European Union — Article 15 accuracy, robustness and cybersecurity for high-risk systems; DORA TLPT where applicable.
    • United Kingdom — Operational resilience testing expected to cover AI-dependent important business services.
    • Gulf sovereign — SAMA cyber framework testing cadence applies to the AI stack as a technology asset.

    AI Cyber & Adversarial

    CYB-27

    Model and weight access is privileged access.

    Access to model weights, fine-tuning pipelines and vector stores is treated as privileged, recertified on cycle and monitored for anomalous retrieval volume.

    Method
    Examine · Test

    Evidence
    Access recertification record and anomaly alert sample for the model estate.

    Jurisdiction delta

    • New York State — Part 500.7 access privilege management, with 500.12 multifactor requirements.
    • European Union — DORA ICT access management for critical or important functions.
    • International standards — ISO/IEC 42001 Annex A control on AI system resources.

    AI Cyber & Adversarial

    TPR-21

    Foundation-model providers are assessed as critical third parties.

    Model and API providers are risk-assessed, contractually bound on change notification and audit access, and assessed for substitutability before dependency deepens.

    Method
    Examine · Interview

    Evidence
    Provider assessment with contract clause map and substitutability rating.

    Jurisdiction delta

    • European Union — DORA critical ICT third-party provisions plus AI Act value-chain obligations under Article 25.
    • Canada — OSFI B-10 material arrangement treatment with concentration analysis.
    • United States (prudential) — Interagency third-party guidance applies across the full relationship lifecycle.
    • United Kingdom — PRA critical third-party regime where the provider is designated.

    AI Third-Party & Supply Chain

    TPR-26

    Silent model change is detected.

    Provider-side model version changes are detected through fingerprinting or contractual notification, and trigger revalidation before continued reliance.

    Method
    Test · Examine

    Evidence
    Version-change detection log with the revalidation that followed.

    Jurisdiction delta

    • United States (prudential) — An undisclosed version change is a model change and reopens validation obligations.
    • European Union — Article 25 obligations shift where substantial modification occurs.
    • Singapore — MAS expects reliance on external models be re-evidenced after change.

    AI Third-Party & Supply Chain

    RES-21

    AI dependency sits inside the impact tolerance, not beside it.

    Important business services that depend on AI carry an impact tolerance covering AI unavailability and degraded-quality output, with a tested manual or fallback path.

    Method
    Examine · Test

    Evidence
    Impact tolerance statement plus fallback test evidence with elapsed recovery time.

    Jurisdiction delta

    • European Union — DORA impact tolerance and scenario testing include ICT third-party AI dependencies.
    • United Kingdom — PRA/FCA operational resilience impact tolerances apply where the service is important.
    • Canada — OSFI E-21 operational resilience expectations extend to AI-dependent processes.
    • United States (markets) — Reg SCI availability standards apply where the system is SCI-covered.

    AI Resilience & Continuity

    RES-27

    Concentration on a single model provider is a board-visible position.

    Provider and model concentration is measured across the estate and reported to the board with the exit horizon for each material dependency.

    Method
    Examine

    Evidence
    Concentration report with exit horizon per provider and board minute.

    Jurisdiction delta

    • European Union — DORA concentration risk assessment for critical ICT providers.
    • Canada — B-10 concentration considerations at enterprise level.
    • United Kingdom — Sectoral concentration is a supervisory question, not only a firm-level one.

    AI Resilience & Continuity

    DSC-02

    Disclosure obligations are met per jurisdiction, from one record.

    Customer, market and supervisory disclosure obligations arising from AI use are mapped per jurisdiction and satisfied from a single underlying evidence record.

    Method
    Examine

    Evidence
    Disclosure matrix per jurisdiction with the evidence artefact satisfying each entry.

    Jurisdiction delta

    • European Union — Article 50 transparency to natural persons; Article 53 for general-purpose model providers.
    • United States (markets) — Disclosure of material AI use and the associated risk factors in filings.
    • Japan — Disclosure in Japanese to the parent board and, where applicable, to JFSA.
    • Singapore — FEAT transparency to affected customers on material AI-driven decisions.
    • United Kingdom — Consumer Duty communications standards apply to the disclosure itself.

    AI Disclosure & Evidence

    DSC-07

    The evidence set is supervisor-ready without a project.

    AI control evidence is continuously collected and assembled into a scoped, time-stamped pack that can be produced on request without a remediation exercise first.

    Method
    Examine · Test

    Evidence
    Generated evidence pack with scope, period and integrity hash.

    Jurisdiction delta

    • European Union — Article 18 documentation and Article 19 automatically generated log retention.
    • United States (prudential) — Examination readiness: evidence produced within the examiner's request window.
    • United States (markets) — Production must be complete and consistent with books-and-records obligations.
    • Gulf sovereign — Evidence must be producible in-jurisdiction where residency constraints apply.

    AI Disclosure & Evidence

    DSC-11

    Incidents involving AI are reported on the correct clock.

    AI-related incidents are triaged against every applicable reporting clock simultaneously, with the shortest clock governing the notification path.

    Method
    Examine · Test

    Evidence
    Incident record with per-regime clock assessment and notification timestamps.

    Jurisdiction delta

    • European Union — AI Act serious incident reporting alongside DORA and NIS2 clocks.
    • New York State — Part 500.17 seventy-two-hour notification to the Superintendent.
    • United States (markets) — Item 1.05 material cybersecurity incident disclosure within four business days.
    • Japan — JFSA and PPC notification paths run in parallel with differing thresholds.
    • Canada — OSFI technology and cyber incident reporting within twenty-four hours.

    AI Disclosure & Evidence

    INF-01

    The AI factory is a declared assurance subject, not an assumed substrate.

    Every AI factory operated or consumed by the institution is registered as an assurance subject with its clusters, model estate, inference paths, tenants and dependencies enumerated.

    Method
    Examine · Interview

    Evidence
    AI factory registration record reconciled against the model inventory and the infrastructure dependency map.

    Jurisdiction delta

    • European Union — DORA ICT asset management extends to the compute estate supporting critical or important functions.
    • United States (prudential) — The supervisory expectation of a complete inventory covers the infrastructure on which models run.
    • Gulf sovereign — Sovereign AI factory registration where compute is state-sponsored or state-adjacent.

    AI Infrastructure & Factory

    INF-04

    Inference locality is known and enforced.

    For each material model, the regions in which inference may execute are declared, enforced at the routing layer and evidenced, with deviation treated as a control event rather than an operational footnote.

    Method
    Examine · Test

    Evidence
    Routing policy with enforcement test showing a disallowed region refused, plus execution-region logs for a sampled period.

    Jurisdiction delta

    • European Union — Chapter V transfer analysis follows the inference region, not the contract.
    • Gulf sovereign — In-country inference may be mandatory for systems touching sovereign data.
    • Japan — APPI third-country transfer disclosure follows where inference actually occurs.
    • New York State — Part 500 asset inventory must reflect where processing of non-public information occurs.

    AI Infrastructure & Factory

    INF-09

    Tenant boundaries inside shared compute are tested, not attested.

    Where the institution shares compute with other tenants, isolation between tenants is tested on a defined cadence and the residual risk of shared infrastructure is accepted at the correct level.

    Method
    Test · Examine

    Evidence
    Tenant isolation test results with scope, cadence and the acceptance record for residual exposure.

    Jurisdiction delta

    • European Union — DORA risk management covers shared and multi-tenant ICT environments supporting important functions.
    • New York State — Part 500.3 risk assessment must address risks from shared infrastructure.
    • Canada — OSFI B-13 technology risk management expects isolation assurance in shared environments.

    AI Infrastructure & Factory

    DCA-01

    Every hosting site is inside the evidence perimeter.

    Each data centre or hosting region on which the AI estate depends is recorded with its jurisdiction, operator, dependencies and evidence locality, so an examiner can reach the evidence without crossing a border unlawfully.

    Method
    Examine

    Evidence
    Hosting register reconciled to the model inventory, with evidence-locality mapping per site.

    Jurisdiction delta

    • European Union — DORA register of information must reflect the data centres supporting critical functions.
    • Canada — OSFI B-10 requires the location of subcontracted operations be known and assessed.
    • Gulf sovereign — Evidence must be producible in-jurisdiction where residency applies.

    Data Centre & Hosting

    DCA-05

    Physical and cloud dependencies are failure-modelled together.

    Power, connectivity, cooling and cloud-region dependencies beneath the AI estate are modelled as a single failure surface, with the failure modes feeding the resilience position rather than sitting in a facilities register.

    Method
    Examine · Test

    Evidence
    Dependency failure model with a worked scenario showing the propagation path into an important business service.

    Jurisdiction delta

    • United Kingdom — Operational resilience requires mapping the resources, including infrastructure, behind each important business service.
    • European Union — DORA scenario testing includes infrastructure dependency failure.
    • Canada — E-21 mapping expectations extend to underlying facilities and cloud dependencies.

    Data Centre & Hosting

    DCA-09

    Provider concentration is measured at the infrastructure layer.

    Concentration on cloud regions, data centre operators and model providers is measured jointly, because a region failure and a provider failure can be the same failure.

    Method
    Examine

    Evidence
    Joint concentration analysis across compute, hosting and model providers with exit horizons.

    Jurisdiction delta

    • European Union — DORA concentration assessment covers ICT providers jointly, not category by category.
    • United States (prudential) — Interagency guidance expects concentration analysis across the third-party inventory.
    • United Kingdom — Critical third-party designation analysis begins from concentration measurement.

    Data Centre & Hosting

    AID-01

    Every agent has a governed identity before it has a capability.

    No agent acts inside the estate without a registered identity carrying its model, version, owner, purpose and objective, recorded before any credential is issued to it.

    Method
    Examine · Test

    Evidence
    Agent identity registry extract with a test showing an unregistered agent denied a credential.

    Jurisdiction delta

    • United States (markets) — Activity initiated by an unidentified agent cannot satisfy supervisory obligations under FINRA 3110.
    • European Union — Article 14 oversight presumes the overseen system is identifiable and bounded.
    • International standards — ISO/IEC 42001 resource management extends to agentic resources acting on the organisation's behalf.

    Agent Identity & Delegation

    AID-04

    Delegation narrows authority, never widens it.

    Where an agent delegates to a child agent, the child inherits a strictly narrower authority envelope, and the delegation is recorded in the lineage graph with the accountable principal unchanged.

    Method
    Examine · Test

    Evidence
    Lineage graph extract with a delegation test showing an inherited envelope exceeding the parent refused.

    Jurisdiction delta

    • United Kingdom — SMCR accountability cannot be delegated; delegation chains must preserve the accountable Senior Manager.
    • United States (prudential) — Each agent in a chain is an in-scope model with its own inventory entry.
    • European Union — Deployer accountability persists across all intermediating systems.

    Agent Identity & Delegation

    AID-08

    Agent termination is a controlled end-state.

    Agent decommissioning revokes credentials, closes open obligations, preserves evidence and updates the lineage graph, with a defined state between active and terminated in which the agent can observe but not act.

    Method
    Examine · Test

    Evidence
    Termination record with credential revocation timestamps and preserved evidence references.

    Jurisdiction delta

    • United States (markets) — Books-and-records obligations survive the termination of the agent that generated them.
    • European Union — Article 18 retention applies to the evidence an agent produced while active.
    • Canada — E-23 model retirement expectations apply to agentic systems.

    Agent Identity & Delegation

    RTM-01

    Policy is enforced at the point of action, not asserted after it.

    Every consequential agent or model action passes a policy enforcement point that evaluates the action against the current authority, context and control state before execution.

    Method
    Examine · Test

    Evidence
    Enforcement point configuration with a test showing a disallowed action blocked before execution.

    Jurisdiction delta

    • European Union — Article 14 oversight measures must be built into the system's operation.
    • United States (prudential) — Preventive controls are examined as part of effective challenge over model use.
    • New York State — Part 500 expects controls that prevent, not merely detect, unauthorised activity.

    Runtime & Gateway Assurance

    RTM-05

    Out-of-envelope actions generate evidence, not silence.

    Any action attempted outside an agent's declared envelope is refused, recorded with full context and routed to the accountable authority with a recommended disposition.

    Method
    Test · Examine

    Evidence
    Out-of-envelope event log with the refusal record, context and the human disposition that followed.

    Jurisdiction delta

    • European Union — Article 12 logging extends to refused and anomalous operation.
    • Singapore — MAS expects anomalies in autonomous operation be escalated with auditability.
    • United States (markets) — Attempted unauthorised activity is within supervisory examination scope.

    Runtime & Gateway Assurance

    RTM-09

    Gateway, mesh and tool boundaries carry one policy, not three.

    AI gateways, MCP servers, service mesh rules and tool interfaces are governed as one policy surface so an action cannot be permitted at one boundary and prohibited at another.

    Method
    Examine · Test

    Evidence
    Unified policy definition with consistency test across at least two boundary types.

    Jurisdiction delta

    • International standards — ISO/IEC 42001 operational control expects consistency across the AI management system.
    • New York State — Part 500.3 policy coverage must extend across the systems in scope without gaps.
    • European Union — DORA requires a coherent, documented ICT risk framework rather than layered fragments.

    Runtime & Gateway Assurance

    EVL-01

    Evaluation is evidence, and is governed as such.

    Every model evaluation, including peer and reciprocal evaluation, produces a structured evidence record with the evaluator, protocol version, domains assessed and evidence state, retained in the corpus.

    Method
    Examine

    Evidence
    Evaluation evidence record with domain coverage and evidence state per domain.

    Jurisdiction delta

    • United States (prudential) — Evaluation outputs that inform reliance are within validation governance under SR 11-7.
    • European Union — Conformity and post-market evidence must be traceable to the assessment that produced it.
    • International standards — ISO/IEC 42001 performance evaluation clause applies to the evaluation process itself.

    Peer Evaluation & Exchange Integrity

    EVL-05

    Evaluators are challenged, and their integrity is assured.

    Peer and reciprocal evaluations are subject to independent challenge, with evaluator conflicts, capability limits and evaluation drift assessed before the evidence is admitted.

    Method
    Examine · Interview

    Evidence
    Evaluator integrity assessment with conflict declaration and challenge record.

    Jurisdiction delta

    • United States (prudential) — Independence of the evaluating party mirrors SR 11-7 validation independence.
    • United Kingdom — SS1/23 expects the validating function to have standing to challenge the evaluated party.
    • European Union — Article 17 quality management covers the integrity of assessment inputs.

    Peer Evaluation & Exchange Integrity

    EVL-09

    Consequential findings are adjudicated by human authority.

    No consequential evaluation finding, including a finding from model-to-model evaluation, changes an assurance state without adjudication by an accountable human authority.

    Method
    Examine · Test

    Evidence
    Adjudication record naming the human authority, the finding and the disposition applied.

    Jurisdiction delta

    • United Kingdom — SMCR places the consequential judgment with the accountable Senior Manager.
    • United States (prudential) — Effective challenge must be human and demonstrable.
    • European Union — Article 14 human oversight cannot be discharged to another model.

    Peer Evaluation & Exchange Integrity

    BHV-01

    Selection risk is read from the whole context, not the model alone.

    Behavioural risk is assessed across the model, its objective, incentives, environment, permissions, feedback loops and the other agents it interacts with, as a single composite read.

    Method
    Examine · Interview

    Evidence
    Behavioural assessment record covering all seven inputs with the composite conclusion and reviewer.

    Jurisdiction delta

    • United States (markets) — Conflicts and incentives in automated systems are within examination scope.
    • European Union — High-risk classification considers the system's purpose and context of use, not its architecture.
    • Singapore — FEAT accountability expects incentive structures around automated decisions be examined.

    Behavioural & Selection Risk

    BHV-04

    Feedback loops are bounded and observed.

    Where an agent's output feeds its own future context or another agent's objectives, the loop is documented, bounded and monitored for amplification or drift.

    Method
    Examine · Test

    Evidence
    Feedback loop register with bounds, monitoring cadence and a drift observation sample.

    Jurisdiction delta

    • United States (prudential) — Model risk includes risks arising from the interaction of models.
    • European Union — Post-market monitoring under Article 72 covers emergent behaviour in operation.
    • United Kingdom — Consumer Duty outcome monitoring applies to self-reinforcing decision patterns.

    Behavioural & Selection Risk

    BHV-07

    Behavioural findings feed the resilience position, with limits stated.

    Behavioural assessment results propagate into the relationship graph and the resilience position, and are never presented as predictions of future behaviour.

    Method
    Examine

    Evidence
    Propagation trace from a behavioural finding to the resilience position, with the stated limits of the assessment.

    Jurisdiction delta

    • United States (prudential) — Supervisory reporting must not present assessments as forecasts of behaviour.
    • European Union — Disclosure of AI capability must not overstate what the assessment establishes.
    • International standards — ISO/IEC 23894 requires the limits of risk assessment be documented.

    Behavioural & Selection Risk

    IPL-01

    The institution can answer where the data went.

    For each material data flow into a model or agent, the institution records source, purpose, permissions, processing location, training use, third parties reached and the contractual and regulatory restrictions that travel with the data.

    Method
    Examine · Test

    Evidence
    Data lineage record traced end to end for a sampled flow, including training-use status.

    Jurisdiction delta

    • European Union — GDPR purpose limitation and Article 10 data governance apply jointly to training use.
    • Japan — APPI purpose-of-use constraints follow the data into model training.
    • New York State — Part 500 data inventory extends to non-public information consumed by AI systems.
    • Gulf sovereign — Residency and sovereignty constraints may prohibit training use entirely.

    AI Data & IP Lineage

    IPL-05

    Training use is a permissioned state, not a default.

    Whether data may be used for training or fine-tuning is an explicit recorded permission per flow, enforced contractually and verified against the provider's actual practice where testable.

    Method
    Examine · Test

    Evidence
    Training-permission register reconciled to contract terms and, where testable, provider configuration.

    Jurisdiction delta

    • European Union — Training use beyond the stated purpose requires its own lawful basis.
    • Canada — OSFI B-10 expects contractual control over subcontractor use of institutional data.
    • United States (markets) — Undisclosed training use of client information raises Reg S-P exposure.

    AI Data & IP Lineage

    IPL-09

    Intellectual property exposure is mapped in both directions.

    The institution records which of its intellectual property enters external models and which third-party intellectual property may be embedded in model output it relies on, with the legal exposure assessed for each direction.

    Method
    Examine · Interview

    Evidence
    Bidirectional IP exposure register with legal assessment per material entry.

    Jurisdiction delta

    • European Union — AI Act transparency obligations interact with text and data mining exceptions under the DSM Directive.
    • United Kingdom — Passing off and database right considerations apply to model output reuse.
    • United States (markets) — Material IP exposure arising from AI use may be a disclosable risk factor.

    AI Data & IP Lineage

    SOV-01

    Sovereignty is treated as a legal relationship, not a server location.

    Each model, agent and data flow carries a sovereignty profile recording which jurisdictions' laws attach, on what basis, and which foreign dependencies could reach the asset.

    Method
    Examine · Interview

    Evidence
    Sovereignty profile per material asset with the legal basis for each jurisdictional attachment.

    Jurisdiction delta

    • European Union — GDPR Article 3 extraterritorial scope and Chapter V together define the attachment.
    • Gulf sovereign — Sovereign data classes carry explicit jurisdictional attachment by instrument.
    • US — Extraterritorial reach of US lawful access instruments is assessed as a foreign dependency for non-US operations.

    Sovereignty & Transnational Operation

    SOV-05

    Transnational operation follows the highest applicable standard.

    Where obligations from multiple jurisdictions intersect on the same asset or action, the highest applicable standard governs, and the intersection analysis is recorded with its rationale.

    Method
    Examine

    Evidence
    Intersection analysis per material transnational flow showing the governing standard and why.

    Jurisdiction delta

    • European Union — GDPR adequacy and transfer analysis require documented comparison of regimes.
    • Canada — OSFI expects cross-border arrangements meet the stricter of the applicable expectations.
    • Japan — APPI requires the foreign regime's equivalence be assessed before transfer.

    Sovereignty & Transnational Operation

    SOV-09

    Jurisdictional change reopens the assurance position.

    A change in applicable law, hosting region or provider dependency automatically reopens the affected assurance states, with the invalidating condition recorded and the reassessment routed to the accountable authority.

    Method
    Examine · Test

    Evidence
    Invalidation event record with the reopened assurance states and the reassessment that followed.

    Jurisdiction delta

    • United States (prudential) — Change in legal environment is a model risk trigger under ongoing monitoring.
    • European Union — Regulatory change affecting a high-risk system requires conformity reassessment.
    • United Kingdom — SMCR accountability includes keeping the control position current as obligations change.

    Sovereignty & Transnational Operation

    This is a published sample. Full test procedures, evidence specifications and the effectiveness rubric behind each control are institutional IP, released under engagement.

    Request the full AI control set
    Institutional control library AI model inventory Control architecture Vertical AI assurance packs
    Cabier Consulting
    Cabier Consulting
    TRUST.INFRASTRUCTURE.FOR.REGULATED.MARKETS

    CABIER GLOBAL ASSURANCE

    Trust infrastructure for AI-native institutions and high-consequence systems.

    YOU CHOOSE THE AI. CABIER ORCHESTRATES THE TRUST.

    Solutions

    • Private Sector Platform
    • Government Platform
    • Government Preview
    • Demo Gallery
    • Schedule Demo

    HUBs

    • Regulatory HUB
    • Legal HUB
    • Policy HUB
    • Crypto HUB
    • Insurance HUB
    • Healthcare HUB
    • Government HUB
    • All HUBs

    Free Tools

    • Regulatory Calendar
    • Framework Calculator
    • Cross-Border Calculator
    • Crisis Stress Tester
    • Control Dependencies
    • Maturity Quiz
    • All Free Tools

    Platform

    • GRCiQ™
    • Managed Cyber GRC
    • Filing Assistant
    • Certification Readiness
    • AI Model Registry
    • ESG Control Center

    Academy

    • Learning Paths
    • Course Catalog
    • Certification Prep
    • Practice Labs
    • All Courses

    Company

    • About Us
    • Founder: Dax Philbert
    • Virtual Consulting
    • Traditional Consulting
    • Staff Augmentation
    • Solutions
    • Contact
    • Careers
    Coverage:United StatesEuropean UnionUnited KingdomCanadaSingaporeAustraliaUAEHong Kong

    © 2026 Cabier Consulting. All rights reserved.

    Privacy PolicyTerms of ServiceCookie Policy