The Enterprise Control Index

APPARENS · v7.1 · JULY 2026

i-DEPOT 158508 · BOIP · CC BY-NC-ND 4.0

CTO

Chief Technology Officer

AI technology strategy, platform selection and build/consume posture

CDO

Chief Data Officer

Data governance, regulatory readiness and evidence integrity

CISO

Chief Info Security Officer

AI-specific threat modelling and security controls

CFO

Chief Financial Officer

AI cost governance, ROI validation and CSRD reporting

Enterprise Architect

Cross-layer integration and architecture coherence

DPO

Data Protection Officer

GDPR compliance, DPIA and data subject rights in AI

PROC

Procurement

Third-party AI governance and vendor concentration risk

LEGAL

Legal / Compliance

AI Act operator-role classification and obligation mapping

Layer-1 decidability — Decidable — a source-anchored rule can settle it — Partly decidable — a gate is mechanical, the substance is judgement — Judgement — no honest rule Dot = where the rule/judgement boundary lies. Open a control to see its boundary and EU AI Act phase-in date. Deciding it for your systems is the app.

CONTROL INDEX FORENSIC EXPOSURE

5 Existential

4 Critical

3 Significant

2 Contained

1 Localised

FRAMEWORK

GLOSSARY

BOUNDARY RULES

ARTIFACTS

RUNTIME

CROSSWALK

AGENTIC PACK

REFERENCES

Alignment on Value

L1Strategy & Accountability

Decision rights · Risk appetite · Operator-role classification · IP policy

L2Ethics & Fairness

Fairness metrics · Bias testing · FRIA · Contestability · RAI Board

PEOPLE & SKILLS

L3

People, Skills &
Operating Model

HUMAN PLANE

Roles · Competencies · Training · Operating model

AI TECHNOLOGY STACK
L4Applications & AgentsUSER PLANE
Enterprise assistants, domain agents and agentic AI application controls
L5AI EngineeringPRODUCTION PLANE
Model lifecycle · Prompt engineering · Evaluation · Testing · Production hardening
L6Data, Context & GovernanceKNOWLEDGE PLANE
Vectors · Lineage · Classification · Consent · GDPR · Synthetic data
L7Systems & SourcesINTEGRATION PLANE
Where AI connects to the organisation's source of truth
L8InfrastructurePHYSICAL PLANE
Compute · Network · Storage · Identity · Sovereign cloud

CONTROL & SUPPORT SHIELD

S1GRC

Governance spine · Evidence · Compliance

S2AI Security & IR

Threat defence · Incident response · Zero trust

S3Supply Chain

Vendor governance · Value chain · Concentration risk

S4Observability

Continuous assurance · Drift · Post-market monitoring

S5FinOps

Cost governance · Token economics · Carbon reporting

Glossary of Terms

Formal definitions using ISO vocabulary discipline where available. All terms are used consistently throughout this specification.

Boundary Rules

When two teams could both own a control, who actually does? These rules prevent governance dead zones — the gaps where nobody is accountable and incidents go unowned.

EXTENSION RULE

Layers may be extended with additional components. New components must pass three tests: (1) Does this belong to an existing layer? (2) Does it duplicate an existing component? (3) Is the evidence output defined?

MOVEMENT RULE

Components may not move between layers without a documented rationale and a boundary rule update. The framework version number increments when boundary rules change.

SCOPE, LIMITATIONS & DISCLAIMERS

IN SCOPE

Enterprise AI governance for organisations that develop, deploy, or consume AI systems
Full AI lifecycle: strategy, design, development, deployment, monitoring, decommissioning
Both build posture (training, fine-tuning, custom models) and consume posture (APIs, SaaS, vendor models)
All AI system types: predictive, generative, agentic, multi-agent
Aligned with EU AI Act, ISO/IEC 42001, NIST AI RMF, OWASP LLM + Agentic Top 10
Technology-agnostic: framework applies regardless of model provider, cloud platform, or deployment pattern

OUT OF SCOPE

AI research governance (pre-commercialisation, lab environments, academic research)
Clinical trial AI governance (covered by Medical Devices Regulation MDR / IVDR)
Military and defence AI governance (covered by national security frameworks)
Sector-specific AI regulations beyond the EU AI Act horizontal framework (e.g., financial services prudential requirements, telecommunications-specific rules)
Implementation services, system integration, or technical deployment guidance
AI model selection, benchmarking, or technical performance optimisation

LIMITATIONS & DISCLAIMERS

Not legal advice. This framework provides governance architecture. It does not substitute for legal advice on jurisdiction-specific obligations. Organisations must obtain qualified legal counsel for compliance determinations.

Not an audit. The maturity self-assessment provides an indicative view of implementation status. It is not a substitute for formal internal audit, external certification (ISO/IEC 42001 via ISO/IEC 42006), or regulatory conformity assessment.

Regulatory landscape evolving. The EU AI Act Omnibus (7 May 2026 political agreement) may change high-risk timelines. CEN-CENELEC harmonised standards are in development. The Legal Register (S1) exists specifically to track these transitions.

Framework, not product. Apparens publishes the framework (CC BY-NC-ND 4.0), writes the book, and conducts adversarial review. The organisation implements. The auditor audits. The regulator regulates. These roles are deliberately separated.

Mandatory Artifacts

Without defined artifacts, governance is aspiration. Eight mandatory outputs — click any for the full specification including trigger, contents, regulatory basis, gates, and evidence.

Runtime Oversight

Every layer and shield control is design-time or point-in-time. This plane adds the controls that ask whether the oversight loop actually closes for agentic AI. Three families: enforcement, sensing, and client-side typing.

Standards Crosswalk

Clause-level mappings from four primary standards to framework layers, shields, and evidence artifacts. These mappings substantiate alignment claims with traceability — not "aligned with" claims, but article-to-control-to-evidence chains.

ISO/IEC 42001

NIST AI RMF

EU AI Act

Related Standards

ISO/IEC 42001:2023 — Requirements for an AI Management System (AIMS). Plan-Do-Check-Act structure mapped to framework layers and shields.

P Plan

Strategy (L1): Policy, context, leadership (§4, §5)
GRC (S1): Risk assessment process (§6.1)
Ethics (L2): AI impact assessment (§6.1.4)
GRC (S1): Objectives & planning (§6.2)
Supply Chain (S3): Supply chain planning (§6.1.5)
Strategy (L1): AI Actor Classification + Legal Register

D Do

Production (L5): AI system development controls (§8)
Knowledge (L6): Data governance, synthetic data (§8.4)
People (L3): Competence & awareness (§7.2, §7.3)
User Plane (L2): System deployment, agent declarations (§8.6)
Security (S2): Security controls, multi-agent (§8.5)

C Check

Observability (S4): Performance evaluation, PMM (§9.1)
GRC (S1): Internal audit (§9.2)
GRC (S1): Management review (§9.3)
ART-04: Ongoing eval cadence report
Ethics (L2): Fairness monitoring handshake with S4

A Act

GRC (S1): Nonconformity & corrective action (§10.1)
Security (S2) + ART-08: Incident learning (§10.2)
GRC (S1): Continual improvement (§10.3)
People (L3): Capability uplift post-incident

Clause	Requirement	Framework mapping	Evidence
§4.1–4.4	Context, interested parties, scope, AIMS	Strategy (L1) + AI Actor Classification	Policy document, scope statement, AI Actor Register
§5.1–5.3	Leadership, policy, roles & responsibilities	Strategy (L1) + People (L3) RACI	Signed policy, RACI matrix, board resolution
§6.1.2	AI risk assessment process	GRC (S1) Risk & Impact Assessments	ART-06 Risk Acceptance Record
§6.1.4	AI impact assessment (societal, ethical)	Ethics (L2) FRIA + GRC (S1)	DPIA/FRIA report
§6.1.5	Supply chain AI risk	Supply Chain (S3)	ART-07 Supplier Assurance Pack
§7.2–7.3	Competence, awareness, communication	People (L3) AI Literacy + Academies	Training records, competence assessments
§8.4	Data for AI systems	Knowledge (L6) + Synthetic Data Gov.	ART-02 Dataset Datasheet
§8.5	Information security for AI	Security (S2) + Multi-Agent Orchestration	Red team reports, adversarial test results
§8.6	AI system operation	User (L4) + Production (L5)	ART-01 System Card, ART-03 Eval Plan, ART-05 Agent Declaration
§9.1	Monitoring, measurement, evaluation	Observability (S4) + PMM System	ART-04 Ongoing Evaluation Cadence
§9.2	Internal audit	GRC (S1) Evidence Factory	Audit logs, audit report
§10.1–10.3	Nonconformity, corrective action, improvement	Security (S2) IR + GRC (S1)	ART-08 AI Incident Report

NIST AI RMF 1.0 (AI 100-1, Jan 2023) + Generative AI Profile (AI 600-1, Jul 2024). Four core functions: GOVERN, MAP, MEASURE, MANAGE. RAG-specific governance actions from AI 600-1 mapped separately.

Function	Outcome	Framework mapping	Evidence
GOVERN 1	Policies, processes, procedures established	Strategy (L1) + GRC (S1) Policy-as-Code	Governance corpus, risk appetite statement
GOVERN 2	Accountability structures and culture	Strategy (L1) Executive Accountability + People (L3)	RACI, board resolution, capability records
GOVERN 3	Workforce AI risk management culture	People (L3) AI Champions + Ethics (L2) RAI Board	Training records, review board minutes
GOVERN 5	Third-party risk governance	Supply Chain (S3)	ART-07 Supplier Assurance Pack
GOVERN 6	AI risk management policies	GRC (S1) + Ethics (L2) + Legal Register	Policy register, norms documentation, legal register
MAP 1	Context established and understood	Strategy (L1) + GRC (S1) AI System Inventory	System inventory, AI Actor Register
MAP 2	Categorisation of AI risks	Ethics (L2) Fairness Metrics Registry	Metric selection rationale (signed by RAI Board)
MAP 3	AI benefits and risks mapped	GRC (S1) Risk & Impact Assessments	DPIA/FRIA report
MAP 5	Likelihood and severity evaluated	GRC (S1) + Ethics (L2) + Strategy (L1) Risk Appetite	ART-06 Risk Acceptance Record
MEASURE 1	Risk measurement approaches	Ethics (L2) Fairness Metrics + Observability (S4)	ART-03 Evaluation Plan, ART-04 Cadence Report
MEASURE 2	AI systems evaluated for risk	Observability (S4) + Production (L5) Release Gates	ART-03 Release criteria results
MEASURE 4	Post-deployment risk monitored	Observability (S4) Drift + HITL + PMM System	ART-04 Ongoing Evaluation
MANAGE 1	Risks prioritised and treated	GRC (S1) + Security (S2) Controls	Risk register, treatment plans
MANAGE 4	Incidents and errors reported	Security (S2) IR + People (L3) Training	ART-08 AI Incident Report
AI 600-1 Agentic	Autonomous action, tool use, multi-step reasoning	User (L4) ART-05 + S2 Multi-Agent + Agentic Pack	ART-05 Agent Control Declaration
AI 600-1 MS-2.5-005	Verify RAG data provenance and grounding	Knowledge (L6) Context Injection + Data Lineage	RAG permission config, lineage graph
AI 600-1 MG-3.1-003	Re-assess model risks after RAG implementation	GRC (S1) Lifecycle Workflow + Observability (S4)	ART-03 refresh post-RAG deployment
AI 600-1 MG-3.2-002	Document RAG adaptations and configurations	Production (L5) Prompt & RAG Governance	RAG configuration documentation

EU AI Act (Regulation (EU) 2024/1689). Application: (a) Chapters I-II from 2 Feb 2025; (b) GPAI from 2 Aug 2025; (c) Art. 6(2) high-risk from 2 Aug 2026; (d) Art. 6(1) from 2 Aug 2027. AI Act Omnibus (7 May 2026) proposes postponing (c) to Dec 2027, (d) to Aug 2028.

Article	Obligation	Applies to	Framework mapping	Evidence
Art. 3(1), 3(63)	AI system and GPAI model definitions	All actors	Strategy (L1) AI Actor Classification	AI Actor Register with operator role per system
Art. 9	Risk management system for high-risk AI	Provider	GRC (S1) + Ethics (L2) + Observability (S4)	ART-06 Risk Acceptance + ART-04 Cadence
Art. 9(2)(c); Art. 72	Post-market monitoring system	Provider	Observability (S4) PMM System	ART-04 Cadence Report, PMM plan
Art. 10	Data and data governance	Provider	Knowledge (L6) + Synthetic Data Gov.	ART-02 Dataset Datasheet
Art. 11 + Annex IV	Technical documentation	Provider	GRC (S1) Evidence Factory + Production (L5)	ART-01 System Card + documentation set
Art. 12	Record-keeping and logging	Provider	GRC (S1) Evidence Factory	Audit logs, evidence chain
Art. 14	Human oversight measures	Provider + Deployer	Observability (S4) HITL + User (L4) ART-05	ART-05 with human oversight pattern field
Art. 15	Accuracy, robustness, cybersecurity	Provider	Security (S2) + Observability (S4) + Production (L5)	Red team reports, ART-03 Eval Plan
Art. 17	Quality management system	Provider	ISO 42001 AIMS (full PDCA)	Full AIMS evidence set
Art. 25–28	Value chain obligations	All value chain actors	Supply Chain (S3) + Strategy (L1) AI Actor Classification	ART-07 + value chain obligation matrix
Art. 26	Deployer obligations — monitoring, reporting	Deployer	People (L3) + Observability (S4)	ART-04 + People RACI + PMM plan
Art. 27	Fundamental rights impact assessment	Public bodies + certain deployers	Ethics (L2) FRIA + GRC (S1)	FRIA report with DPO + RAI Board sign-off
Art. 49	EU database registration	Provider + Deployer	GRC (S1) AI System Inventory	Registration confirmation
Art. 50	Transparency — AI interaction disclosure	Provider + Deployer	User (L4) Art. 50 Disclosure	Disclosure mechanism evidence
Art. 51–56	GPAI model obligations	GPAI provider	Supply Chain (S3) + GPAI Code of Practice	ART-07 + GPAI compliance documentation
Art. 73	Serious incident reporting (15 days)	Provider + Deployer	Security (S2) IR + Observability (S4) PMM	ART-08 AI Incident Report
Art. 86	Right to explanation for AI decisions	Deployer	Ethics (L2) Contestability & Recourse	Contestability mechanism documentation
Art. 95	GPAI transparency (model cards)	GPAI provider	Supply Chain (S3) + ART-01	ART-01 (requested from vendor)

Additional ISO/IEC, OWASP, and NIST standards mapped to the framework. Status as of June 2026.

Standard	Scope	Framework mapping	Status
ISO/IEC 23894:2023	AI risk management guidance — bridges ISO 31000 to AI-specific risk sources (bias, drift, adversarial inputs). Annex C: lifecycle mapping.	Strategy (L1) risk methodology + GRC (S1)	Published. EN adopted Feb 2024.
ISO/IEC 42005:2025	AI system impact assessment — guidance for assessing impact on individuals and society.	Ethics (L2) + GRC (S1) FRIA	Published May 2025.
ISO/IEC 42006:2025	Requirements for AI audit and certification bodies — establishes the 42001 certification ecosystem.	GRC (S1) audit & assurance	Published Jul 2025.
ISO/IEC 27001:2022 + Amd 1:2024	Information security management systems. Climate Action amendment adds environmental context to risk.	Security (S2) + Infrastructure (L8)	Published. Amd 1 Feb 2024.
ISO 31000:2018	Risk management guidelines. Foundation for framework risk methodology. Edition 3 in CD stage.	Strategy (L1) Risk Appetite + GRC (S1)	Published. Ed. 3 in CD (2026).
ISO/IEC 27090	AI cybersecurity — guidance for addressing security threats and failures in AI systems.	Security (S2)	FDIS stage. Expected mid-2026.
ISO/IEC 27091	AI privacy — membership inference, model inversion, lifecycle privacy engineering.	Data (L6) + Security (S2)	DIS stage. Expected end 2026.
OWASP LLM Top 10 v2025.1	Ten critical security risks for LLM applications: prompt injection (LLM01), sensitive info (LLM02), supply chain (LLM03), excessive agency (LLM06), misinformation (LLM09).	Security (S2) + Agentic Pack	Published. 2026 edition in progress.
OWASP Agentic Top 10 (2026)	Ten critical risks for autonomous AI agents (ASI01-ASI10): cascading permissions, impersonation, memory poisoning, cross-agent escalation, emergent behaviour.	Security (S2) Multi-Agent + User (L4)	Published Dec 2025.
NIST AI 600-1	Generative AI Profile for AI RMF. Includes RAG-specific governance actions (MS-2.5-005, MG-3.1-003, MG-3.2-002).	Production (L5) + Knowledge (L6) + User (L4)	Published Jul 2024.
NIST AI Agent Initiative	RFI on AI agent security + concept paper on agent identity & authorisation. Interoperability profile planned Q4 2026.	Security (S2) + User (L4)	RFI Feb 2026. Profile Q4 2026.
NIST IR 8596	Cyber AI Profile — AI cybersecurity guidance using CSF 2.0 framework.	Security (S2)	Draft Dec 2025.
GPAI Code of Practice	Transparency and governance obligations for GPAI model providers. Endorsed by European Commission.	Supply Chain (S3) + Production (L5)	Published Jul 2025. Endorsed Aug 2025.

Agentic AI Control Pack

Agentic and LLM systems introduce risk multipliers not present in conventional software: autonomous multi-step reasoning, tool invocation, memory, and non-deterministic outputs. This pack maps OWASP LLM Top 10 v2025.1 + Agentic Applications 2026 threats to framework controls.

References

Standards, regulations, and academic foundations. All references are cited at clause or article level in the framework controls and crosswalk.

Registered intellectual property · i-DEPOT 158508 (BOIP)

AI Control Index v7.1 · 8 Layers (L1–L8) · 5 Shields (S1–S5) · 109 Controls · 8 Mandatory Artifacts · 10 Boundary Rules · Runtime Oversight plane (RT · SN · TY) · Evidentiary Sufficiency E0–E3 · Layer-1 Decidability map (STRONG / PARTIAL / NONE + EU AI Act phase-in)
© 2026 Jeroen Janssen / Apparens · apparens.nl · CC BY-NC-ND 4.0

The AI Control Index, the Strategic Red Teaming Method, and all associated frameworks, methodologies, and visual designs are the intellectual property of Apparens. Unauthorised reproduction, distribution, or derivative use is prohibited.