Iso42001 Ai Management

ISO 42001 AI Management System

Tools and guidance for ISO/IEC 42001:2023 — the first international standard for AI Management Systems (AIMS).

---

Table of Contents

• Tools
• AIMS Readiness Checker
• AI Impact Assessor
• Reference Guides
• Workflows
• Standard Overview

---

Tools

AIMS Readiness Checker

Assesses organizational readiness against all ISO 42001 clauses and Annex A controls. Scores each clause on a 0-100 scale and identifies gaps for certification preparation.

# Assess readiness from a JSON profile
python scripts/aims_readiness_checker.py --input org_profile.json

# Generate a blank input template
python scripts/aims_readiness_checker.py --template > org_profile.json

# JSON output for automation
python scripts/aims_readiness_checker.py --input org_profile.json --json

# Export report to file
python scripts/aims_readiness_checker.py --input org_profile.json --output report.json

Assessment Areas:

Output:

• Overall readiness score (0-100)
• Per-clause scores with maturity level (Initial/Developing/Defined/Managed/Optimized)
• Annex A control implementation status (Implemented/Partial/Not Implemented/Not Applicable)
• Gap analysis with prioritized recommendations
• Certification readiness assessment (Ready/Near Ready/Significant Gaps)

---

AI Impact Assessor

Generates comprehensive AI impact assessments evaluating fairness, transparency, safety, privacy, and security dimensions. Maps impacts to interested parties and provides risk treatment recommendations.

# Assess an AI system from a JSON description
python scripts/ai_impact_assessor.py --input ai_system.json

# Generate a blank input template
python scripts/ai_impact_assessor.py --template > ai_system.json

# Export assessment report
python scripts/ai_impact_assessor.py --input ai_system.json --output assessment.json

# Generate markdown report
python scripts/ai_impact_assessor.py --input ai_system.json --format markdown --output assessment.md

Assessment Dimensions:

Features:

• Risk scoring per dimension (Low/Medium/High/Critical)
• Interested party impact mapping (users, affected individuals, society, regulators)
• Risk treatment options (Avoid, Mitigate, Transfer, Accept)
• Regulatory mapping (EU AI Act risk tier, ISO 42001 Annex A controls)
• Residual risk calculation after treatment
• Markdown and JSON report generation

---

Reference Guides

ISO 42001 Clause Guide

references/iso42001-clause-guide.md

Comprehensive clause-by-clause guidance:

• All clauses (4-10) with requirements and implementation steps
• Annex A controls (A.2-A.10) detailed with evidence requirements
• Audit questions per clause for internal audit preparation
• Common nonconformity findings and how to avoid them
• Required documented information per clause
• Cross-references to ISO 27001, ISO 9001, and EU AI Act

AI Lifecycle Management

references/ai-lifecycle-management.md

End-to-end AI system lifecycle guidance:

• Lifecycle stages: design, development, testing, deployment, monitoring, retirement
• Design and development controls (requirements, architecture, coding standards)
• Testing and validation requirements (functional, bias, robustness, performance)
• Deployment procedures (staging, canary, rollback, approval gates)
• Monitoring and maintenance (drift detection, performance degradation, retraining)
• Retirement and decommissioning (data disposal, model archival, stakeholder notification)
• Data management across lifecycle (quality, provenance, bias assessment, lineage)
• Model versioning and change management (version control, change impact, approval workflows)

---

Workflows

Workflow 1: ISO 42001 Readiness Assessment

Step 1: Define AIMS scope
        → Identify AI systems in scope
        → Determine organizational boundaries
        → Document interested parties and requirements

Step 2: Generate assessment template
        → python scripts/aims_readiness_checker.py --template > org_profile.json
        → Fill in organizational details and current state

Step 3: Run readiness assessment
        → python scripts/aims_readiness_checker.py --input org_profile.json

Step 4: Review results
        → Address critical gaps (Clauses 5, 6, 8 typically weakest)
        → Prioritize Annex A controls by risk
        → Develop remediation roadmap

Step 5: Conduct AI impact assessments
        → python scripts/ai_impact_assessor.py --template > ai_system.json
        → Assess each in-scope AI system
        → python scripts/ai_impact_assessor.py --input ai_system.json

Step 6: Plan implementation
        → See references/iso42001-clause-guide.md for requirements
        → See references/ai-lifecycle-management.md for operational controls

Workflow 2: AI System Impact Assessment

Step 1: Identify AI system for assessment
        → Document system purpose, inputs, outputs, and decisions
        → Identify affected individuals and groups

Step 2: Generate assessment template
        → python scripts/ai_impact_assessor.py --template > ai_system.json
        → Complete all sections (model details, data sources, deployment context)

Step 3: Conduct assessment
        → python scripts/ai_impact_assessor.py --input ai_system.json --format markdown --output report.md

Step 4: Review dimension scores
        → Fairness: check for bias in training data and outcomes
        → Transparency: verify explainability mechanisms
        → Safety: validate failure modes and human oversight
        → Privacy: confirm data protection measures
        → Security: assess adversarial resilience

Step 5: Implement risk treatments
        → Apply recommended mitigations per dimension
        → Document residual risk acceptance decisions
        → Assign treatment owners and timelines

Step 6: Monitor and review
        → Schedule periodic reassessment (quarterly minimum)
        → Track treatment implementation progress
        → Update assessment when system changes materially

Workflow 3: AIMS Certification Preparation

Step 1: Gap analysis
        → python scripts/aims_readiness_checker.py --input org_profile.json
        → Target overall score of 80+ for certification readiness

Step 2: Document AIMS
        → AI policy (Clause 5.2)
        → AIMS scope (Clause 4.3)
        → Risk assessment methodology (Clause 6.1)
        → Statement of Applicability for Annex A controls
        → AI objectives (Clause 6.2)

Step 3: Implement operational controls
        → AI lifecycle procedures (Clause 8)
        → Data management processes (Annex A.7)
        → Third-party management (Annex A.10)
        → Impact assessments for all AI systems (Annex A.5)

Step 4: Conduct internal audit
        → Use references/iso42001-clause-guide.md audit questions
        → Document findings and corrective actions
        → Verify closure of nonconformities

Step 5: Management review
        → Present AIMS performance to top management
        → Review AI objectives achievement
        → Obtain commitment for continual improvement

Step 6: Stage 1 and Stage 2 audits
        → Stage 1: Documentation review (readiness check)
        → Stage 2: Implementation effectiveness audit
        → Address any nonconformities from audit

---

Standard Overview

ISO 42001:2023 Overview

ISO/IEC 42001:2023 is the world's first international standard for AI Management Systems (AIMS). Published in December 2023, it provides a framework for organizations to responsibly develop, provide, and use AI systems. The standard follows the ISO Harmonized Structure (Annex SL) for management system standards, enabling integration with ISO 27001, ISO 9001, and ISO 14001.

Key Characteristics:

• Certifiable management system standard
• Technology-neutral (applies to any AI approach)
• Risk-based approach to AI governance
• PDCA (Plan-Do-Check-Act) cycle
• Applicable to organizations of any size and sector

AIMS Framework (Plan-Do-Check-Act)

Context of the Organization (Clause 4)

Leadership (Clause 5)

AI Policy Must Include:

• Commitment to responsible AI development and use
• Ethical principles guiding AI decisions
• Alignment with applicable legal and regulatory requirements
• Commitment to continual improvement of the AIMS
• Framework for setting AI objectives

AI Governance Structure:

• AI governance board or committee
• AI system owners with defined accountability
• Data stewards for AI data management
• Ethics review function
• Incident response roles

Planning (Clause 6)

AI Risk Assessment Must Cover:

• Fairness and non-discrimination risks
• Transparency and explainability gaps
• Safety and reliability concerns
• Privacy and data protection risks
• Security vulnerabilities
• Accountability gaps
• Societal and environmental impacts

Support (Clause 7)

Operation (Clause 8)

AI System Lifecycle Stages: 1. Design: Requirements, architecture, ethical review 2. Development: Data preparation, model training, coding standards 3. Testing: Functional, bias, robustness, performance validation 4. Deployment: Staging, approval, monitoring setup 5. Operation: Performance monitoring, drift detection, incident response 6. Retirement: Decommissioning, data disposal, stakeholder notification

Data Management for AI:

• Data quality assessment and improvement
• Data provenance and lineage tracking
• Bias assessment in training and evaluation data
• Data governance and access controls
• Personal data protection measures
• Data retention and disposal procedures

Third-Party and Supplier Management:

• AI component supplier evaluation
• Third-party AI service agreements
• Supply chain risk assessment
• Ongoing supplier monitoring

Performance Evaluation (Clause 9)

AI Performance Metrics:

• Model accuracy, precision, recall
• Fairness metrics (demographic parity, equalized odds)
• Latency and availability
• Drift indicators (data drift, concept drift)
• Incident frequency and severity
• Consumer complaint rates

Improvement (Clause 10)

Annex A Controls

Annex B — Implementation Guidance

Annex B provides non-normative guidance for implementing Annex A controls:

• Practical examples for each control objective
• Scalability guidance for different organization sizes
• Sector-specific considerations
• Integration points with existing management systems

Annex C — AI Risk Sources and Objectives

AI-specific risk sources organized by category:

• Technical risks: Model failure, data quality, adversarial attacks, drift
• Ethical risks: Bias, discrimination, lack of transparency, autonomy erosion
• Legal risks: Regulatory non-compliance, liability, intellectual property
• Societal risks: Job displacement, misinformation, environmental impact
• Organizational risks: Skill gaps, dependency, reputation damage

AI-specific control objectives:

• Ensure fairness and non-discrimination
• Maintain transparency and explainability
• Guarantee safety and reliability
• Protect privacy and data
• Secure AI systems against threats
• Enable accountability and governance

Annex D — Use of AIMS Across Domains

Sector-specific considerations:

• Healthcare: Patient safety, clinical validation, regulatory approval (FDA, MDR)
• Finance: Algorithmic trading, credit scoring, anti-money laundering
• Autonomous systems: Safety-critical decisions, human override, fail-safe design
• Human resources: Hiring bias, employee monitoring, fairness
• Public sector: Citizen impact, democratic values, public trust

Relationship to Other Standards

Relationship to EU AI Act

Certification Process

Implementation Roadmap

Phase 1 — Foundation (Months 1-3):

• Define AIMS scope and boundaries
• Establish AI governance structure
• Develop AI policy
• Conduct initial AI system inventory
• Define risk assessment methodology

Phase 2 — Core Implementation (Months 4-6):

• Conduct AI risk assessments for all in-scope systems
• Perform impact assessments (Annex A.5)
• Implement AI lifecycle controls (Annex A.6)
• Establish data management processes (Annex A.7)
• Develop third-party management procedures (Annex A.10)

Phase 3 — Operationalize (Months 7-9):

• Deploy monitoring and measurement (Clause 9.1)
• Train personnel on AIMS roles and responsibilities
• Implement incident management procedures
• Conduct awareness programs for AI literacy
• Establish communication processes

Phase 4 — Verify and Certify (Months 10-12):

• Conduct internal audit (Clause 9.2)
• Hold management review (Clause 9.3)
• Address nonconformities
• Prepare for Stage 1 certification audit
• Compile evidence packages per clause

---

Troubleshooting

---

Success Criteria

• Overall readiness score of 80+ for certification readiness -- as measured by aims_readiness_checker.py, with all clauses at Defined maturity level or above
• AI policy established and communicated -- including ethical principles, responsible AI commitment, legal compliance alignment, continual improvement, and framework for AI objectives, with formal management approval
• AI impact assessments completed for all in-scope AI systems -- covering all six dimensions (fairness, transparency, safety, privacy, security, accountability) with risk treatments documented and residual risk accepted by management
• AI risk assessment methodology covers all required categories -- fairness, transparency, safety, privacy, security, accountability, and societal/environmental impacts, with defined risk criteria and tolerance levels
• Annex A controls implemented with evidence -- A.2 (Policies) through A.10 (Third-Party) with documented procedures, records, and evidence artifacts suitable for certification audit
• Internal audit conducted against all AIMS clauses -- with findings documented, corrective actions tracked to closure, and management review completed with documented improvement decisions
• AI lifecycle procedures operational -- covering design, development, testing, deployment, monitoring, and retirement stages with documented controls at each gate

---

Scope & Limitations

In Scope:

• ISO 42001:2023 readiness assessment across all clauses (4-10) and Annex A controls (A.2-A.10)
• AI impact assessment across six dimensions (fairness, transparency, safety, privacy, security, accountability)
• AIMS certification preparation including gap analysis, implementation roadmap, and audit readiness
• AI lifecycle management guidance (design through retirement)
• Data management for AI systems (quality, provenance, bias, governance)
• Third-party AI supplier management and evaluation
• Regulatory mapping to EU AI Act requirements
• Integration guidance with ISO 27001, ISO 9001, and ISO 14001

Out of Scope:

• Actual AI model development, training, testing, or deployment -- this skill provides governance frameworks, not ML engineering
• Certification body selection, audit scheduling, or certification fee negotiation
• Ethical review board establishment or ethical decision-making beyond procedural guidance
• Specific AI fairness algorithm implementation (e.g., adversarial debiasing, calibrated equalized odds) -- use eu-ai-act-specialist bias detector for technical testing
• Environmental impact measurement or carbon footprint calculation for AI training

Important Notes:

• ISO 42001 certification follows a 3-year cycle with annual surveillance audits at 12-month intervals
• Major certification bodies (BSI, DNV, TUV, LRQA) have operationalized ISO 42001 audit services as of 2025-2026
• Many organizations pursue dual alignment: ISO 42001 certification for governance controls plus EU AI Code of Practice for regulatory expectations
• The standard's Annex SL structure enables direct integration with ISO 27001, reducing redundant documentation and audit effort

---

Integration Points

---

Tool Reference

aims_readiness_checker.py

Assesses organizational readiness against all ISO 42001:2023 clauses and Annex A controls.

Assessment Scope: Clause 4 (Context), Clause 5 (Leadership), Clause 6 (Planning), Clause 7 (Support), Clause 8 (Operation), Clause 9 (Performance), Clause 10 (Improvement), and Annex A controls (A.2-A.10).

Output: Overall readiness score (0-100), per-clause scores with maturity level (Initial/Developing/Defined/Managed/Optimized), Annex A control implementation status, gap analysis with prioritized recommendations, and certification readiness assessment (Ready/Near Ready/Significant Gaps).

ai_impact_assessor.py

Generates comprehensive AI impact assessments across six risk dimensions with regulatory mapping.

Assessment Dimensions: Fairness (bias, discrimination, equity), Transparency (explainability, interpretability), Safety (reliability, robustness, harm prevention), Privacy (data protection, consent, minimization), Security (adversarial resilience, access control), Accountability (governance, responsibility, auditability).

Output: Per-dimension risk scoring (Low/Medium/High/Critical), interested party impact mapping, risk treatment options (Avoid/Mitigate/Transfer/Accept), regulatory mapping (EU AI Act risk tier, ISO 42001 Annex A controls), residual risk calculation, and markdown or JSON report.

AI Lifecycle Management Guide

Comprehensive guidance for managing AI systems through all lifecycle stages per ISO 42001:2023 Clause 8 and Annex A.6. Covers design, development, testing, deployment, monitoring, retirement, data management, and model versioning.

---

Table of Contents

• Lifecycle Overview
• Stage 1: Design
• Stage 2: Development
• Stage 3: Testing and Validation
• Stage 4: Deployment
• Stage 5: Monitoring and Maintenance
• Stage 6: Retirement and Decommissioning
• Data Management Across Lifecycle
• Model Versioning and Change Management

---

Lifecycle Overview

┌──────────┐    ┌─────────────┐    ┌──────────┐    ┌────────────┐
│  DESIGN  │───>│ DEVELOPMENT │───>│ TESTING  │───>│ DEPLOYMENT │
└──────────┘    └─────────────┘    └──────────┘    └────────────┘
     ^                                                    │
     │                                                    v
     │          ┌──────────────┐    ┌────────────────────────────┐
     └──────────│  RETIREMENT  │<───│ MONITORING & MAINTENANCE   │
                └──────────────┘    └────────────────────────────┘

Gate Reviews: Each transition between stages requires a formal gate review to verify readiness, address risks, and approve progression.

---

Stage 1: Design

Requirements Definition

Functional Requirements:

• Problem statement and business objective
• Input data specification (types, sources, formats)
• Output specification (predictions, classifications, recommendations)
• Performance requirements (accuracy, latency, throughput)
• Scalability and availability requirements

Responsible AI Requirements:

• Fairness criteria and protected attribute handling
• Transparency and explainability requirements
• Safety and reliability requirements
• Privacy and data protection requirements
• Security requirements
• Human oversight requirements
• Regulatory compliance requirements (EU AI Act, sector-specific)

Architecture Design

Design Decisions to Document:

• Model architecture selection rationale
• Training approach (supervised, unsupervised, reinforcement, fine-tuning)
• Feature engineering strategy
• Data pipeline architecture
• Inference pipeline design
• Integration architecture (APIs, batch, real-time)
• Failover and fallback design
• Human-in-the-loop integration points

Ethical Review

Before proceeding to development:

• Conduct ethical review with diverse stakeholders
• Assess potential for discrimination or bias
• Evaluate societal impact (positive and negative)
• Consider dual-use risks
• Document ethical review findings and mitigations
• Obtain ethics board approval for high-risk applications

Impact Assessment

Per ISO 42001 Annex A.5:

• Assess impact on all identified interested parties
• Evaluate fairness, transparency, safety, privacy, security dimensions
• Map impacts to risk levels
• Define risk treatment measures
• Document residual risk and acceptance rationale

Gate G1 Checklist:

• [ ] Requirements documented and approved
• [ ] Architecture design documented
• [ ] Ethical review completed
• [ ] Impact assessment completed
• [ ] Resource requirements identified
• [ ] Development plan approved
• [ ] Risks identified and treatment planned

---

Stage 2: Development

Data Preparation

Data Collection:

• Document data sources with provenance
• Verify legal basis for data collection and use
• Assess data representativeness for target population
• Document consent/authorization for personal data
• Establish data labeling guidelines and quality criteria

Data Quality Assessment:

Bias Assessment in Data:

• Analyze demographic distribution in training data
• Identify underrepresented groups
• Check for historical bias in labels
• Assess proxy variables for protected attributes
• Document mitigation steps taken (resampling, reweighting, augmentation)

Model Training

Development Standards:

• Use version control for all code, configurations, and data references
• Follow coding standards appropriate to the ML framework
• Document hyperparameter selection rationale
• Implement reproducibility measures (random seeds, environment pinning)
• Conduct code reviews for ML pipeline code
• Separate training, validation, and test datasets (no data leakage)

Training Documentation:

• Training data description (size, composition, preprocessing)
• Model architecture and configuration
• Hyperparameter search strategy and results
• Training process (epochs, convergence, compute used)
• Validation results during training
• Known limitations discovered during training

Code Review for AI

In addition to standard code review:

• Verify data preprocessing does not introduce bias
• Check for data leakage between train/test splits
• Validate feature engineering logic
• Review model serialization and loading
• Verify reproducibility settings
• Check for hardcoded assumptions or thresholds

Gate G2 Checklist:

• [ ] Training data documented with provenance
• [ ] Data quality assessment completed
• [ ] Bias assessment conducted on data
• [ ] Model trained and validated
• [ ] Code reviewed
• [ ] Baseline performance metrics met
• [ ] Development artifacts versioned

---

Stage 3: Testing and Validation

Functional Testing

Bias and Fairness Testing

Required Metrics:

Testing Approach: 1. Define protected attributes and demographic groups 2. Create test datasets stratified by protected attributes 3. Measure all fairness metrics per group pair 4. Document any disparities and root cause analysis 5. Apply mitigation if thresholds are not met 6. Retest after mitigation

Robustness Testing

• Adversarial testing: Test with adversarial examples (perturbed inputs)
• Out-of-distribution: Test with data outside training distribution
• Noise tolerance: Test with noisy or corrupted inputs
• Missing data: Test with missing features
• Temporal robustness: Test with data from different time periods
• Stress testing: Test under resource constraints (CPU, memory, network)

Security Testing

• Model extraction: Assess vulnerability to model stealing attacks
• Data poisoning: Validate training pipeline integrity
• Prompt injection: Test for prompt injection vulnerabilities (LLMs)
• API security: Standard API security testing (authentication, authorization, rate limiting)
• Input sanitization: Validate input validation effectiveness
• Output filtering: Verify output filtering for harmful content

Validation by Independent Party

For high-risk AI systems:

• Validation by team independent from development
• Use held-out validation dataset not seen during development
• Validate against original requirements and responsible AI criteria
• Document validation results and any residual concerns
• Obtain sign-off from independent validator

Gate G3 Checklist:

• [ ] All functional tests passed
• [ ] Bias and fairness testing completed — thresholds met
• [ ] Robustness testing passed
• [ ] Security testing completed — no critical vulnerabilities
• [ ] Independent validation completed (if required)
• [ ] Test results documented and reviewed
• [ ] Deployment plan approved

---

Stage 4: Deployment

Pre-Deployment

Deployment Plan Must Include:

• Deployment environment and infrastructure
• Deployment approach (canary, blue-green, rolling, big-bang)
• Rollback procedure and triggers
• Monitoring and alerting configuration
• Stakeholder notification plan
• Go/no-go criteria and decision authority

Canary and Staged Deployment

Recommended Approach: 1. Shadow mode: Run new model alongside existing system, compare outputs 2. Canary (1-5%): Route small percentage of traffic to new model 3. Gradual rollout (5-25-50-100%): Increase traffic incrementally 4. Full deployment: Route all traffic after validation at each stage

At each stage, monitor:

• Model performance metrics vs. baseline
• Fairness metrics vs. baseline
• Error rates and failure modes
• Latency and resource utilization
• User feedback and complaints
• Business metrics impact

Approval Gates

Who must approve deployment:

• AI system owner
• Technical lead
• Risk/compliance representative
• Ethics reviewer (for high-risk systems)
• Business stakeholder

Approval documentation:

• Test results summary
• Risk assessment status
• Compliance verification
• Deployment plan
• Rollback plan

Post-Deployment Verification

Within first 24-72 hours:

• Verify monitoring and alerting is operational
• Confirm model performance matches testing results
• Check fairness metrics in production
• Verify logging and audit trail capture
• Confirm rollback capability is functional
• Validate user-facing transparency mechanisms

Gate G4 Checklist:

• [ ] Deployment plan executed
• [ ] Monitoring and alerting confirmed operational
• [ ] Production performance verified against test results
• [ ] Rollback tested and confirmed functional
• [ ] Stakeholders notified
• [ ] Post-deployment verification completed

---

Stage 5: Monitoring and Maintenance

Continuous Monitoring

Performance Monitoring:

• Model accuracy, precision, recall (vs. baseline and threshold)
• Prediction distribution (detect distribution shifts)
• Response latency (P50, P95, P99)
• Throughput and error rates
• Resource utilization (CPU, GPU, memory)

Drift Detection:

Fairness Monitoring:

• Continuous tracking of fairness metrics per demographic group
• Alert when fairness metrics cross thresholds
• Quarterly comprehensive fairness review

Incident Detection:

• Anomaly detection on model outputs
• User complaint/feedback monitoring
• Automated alerting for performance degradation
• Security event monitoring

Maintenance Activities

Routine Maintenance:

• Weekly: Review monitoring dashboards and alerts
• Monthly: Performance report and trend analysis
• Quarterly: Comprehensive fairness and drift review
• Annually: Full model revalidation and impact reassessment

Retraining Triggers:

• Significant data drift detected (PSI > 0.25)
• Performance below acceptable threshold
• Concept drift confirmed by ground truth analysis
• New data sources or features available
• Regulatory or business requirement changes
• Bias detected in production fairness monitoring

Retraining Process: 1. Identify need for retraining (trigger event) 2. Prepare updated training data with quality and bias checks 3. Retrain model following development standards 4. Conduct full testing suite (functional, fairness, robustness, security) 5. Validate against production performance 6. Deploy via staged deployment process 7. Monitor post-retraining performance

Incident Response

Severity Levels:

Response Procedure: 1. Detect: Alert triggered or report received 2. Triage: Classify severity, assign owner 3. Contain: Disable system, rollback, or enable human override 4. Investigate: Root cause analysis 5. Resolve: Fix and verify 6. Communicate: Notify affected parties 7. Learn: Document lessons learned, update controls

---

Stage 6: Retirement and Decommissioning

Retirement Triggers

• Model superseded by improved version
• Business use case no longer relevant
• Regulatory changes making system non-compliant
• Unacceptable risk that cannot be mitigated
• Cost-benefit analysis no longer favorable
• Technology obsolescence

Retirement Planning

Retirement Plan Must Include:

• Retirement timeline and milestones
• Stakeholder notification plan
• Data disposition plan (archive, delete, anonymize)
• Model artifact archival plan
• Documentation preservation
• Service migration plan (for dependent systems)
• User communication and transition support
• Compliance obligations during and after retirement

Data Disposition

Stakeholder Notification

Notify the following parties before retirement:

• End users and affected individuals
• Business stakeholders and system owners
• Downstream system owners (API consumers)
• Regulatory bodies (if required)
• Third-party providers and suppliers
• Internal teams (development, operations, support)

Notification Timeline:

• 90 days: Initial notification of planned retirement
• 30 days: Reminder with migration guidance
• 7 days: Final notification
• Day 0: System retired, redirect/deprecation notice active

Gate G5 Checklist:

• [ ] Retirement plan approved
• [ ] Stakeholders notified (90/30/7 day)
• [ ] Data disposition executed per plan
• [ ] Model artifacts archived
• [ ] Documentation preserved
• [ ] Dependent systems migrated
• [ ] Retirement verified and confirmed

---

Data Management Across Lifecycle

Data Governance Framework

Principles:

• Quality: Data must meet defined quality standards before use in AI
• Provenance: Data origin and transformation history must be traceable
• Protection: Personal and sensitive data must be protected throughout lifecycle
• Minimization: Only data necessary for the purpose should be collected and retained
• Fairness: Data must be assessed for bias and representativeness
• Retention: Data must be retained only as long as necessary

Data Quality Management

Quality Dimensions:

Data Provenance and Lineage

Document for all data used in AI systems:

• Origin: Where did the data come from?
• Collection method: How was it collected?
• Legal basis: What authorizes its use?
• Transformations: What preprocessing was applied?
• Versioning: Which version of the data was used?
• Quality scores: What were the quality assessment results?
• Bias assessment: What bias checks were performed?

Data Bias Assessment

Assessment Steps: 1. Define protected attributes and demographic groups 2. Analyze representation in training data vs. target population 3. Check for historical bias in labels or outcomes 4. Identify proxy variables for protected attributes 5. Test for measurement bias (different accuracy across groups) 6. Document findings and mitigation strategies 7. Repeat assessment when data is updated

Bias Mitigation Techniques:

---

Model Versioning and Change Management

Version Control Requirements

What to Version:

• Source code (ML pipeline, preprocessing, inference)
• Model artifacts (trained weights, architecture files)
• Training data references (dataset version, not data copies)
• Configuration files (hyperparameters, feature definitions)
• Test datasets and evaluation scripts
• Documentation (model card, impact assessment)

Versioning Scheme

Recommended: Semantic Versioning for Models

MAJOR.MINOR.PATCH

MAJOR: Architecture change, new training data source, significant behavior change
MINOR: Hyperparameter tuning, incremental retraining, feature addition
PATCH: Bug fix, configuration update, documentation update

Example:

• v1.0.0: Initial production deployment
• v1.1.0: Retraining with updated data
• v1.1.1: Fixed preprocessing bug
• v2.0.0: New model architecture

Change Management Process

Change Types:

Change Request Process: 1. Submit change request with rationale and impact analysis 2. Assess impact on fairness, safety, privacy, security 3. Review by appropriate authority (see table above) 4. Implement change in development environment 5. Execute required testing 6. Deploy via staged deployment process 7. Verify change in production 8. Update documentation and version records

Model Registry

Maintain a central model registry containing:

• Model ID and version
• Architecture and framework
• Training data version reference
• Performance metrics (training, validation, production)
• Fairness metrics
• Owner and team
• Deployment status (development, staging, production, retired)
• Risk classification
• Dependencies (libraries, infrastructure)
• Approval records
• Known limitations and constraints

Rollback Procedures

When to Roll Back:

• Performance degradation exceeding threshold
• Fairness violation detected
• Security vulnerability discovered
• Unexpected behavior in production
• Business impact exceeding risk appetite

Rollback Process: 1. Decision to rollback (system owner or on-call authority) 2. Redirect traffic to previous model version 3. Verify previous version is serving correctly 4. Disable new version deployment 5. Notify stakeholders of rollback 6. Investigate root cause 7. Plan corrective action before re-deployment 8. Document rollback decision and rationale

ISO 42001:2023 Clause-by-Clause Guide

Comprehensive implementation guidance for ISO/IEC 42001:2023 — AI Management Systems (AIMS). Covers all clauses (4-10), Annex A controls (A.2-A.10), audit questions, common findings, and evidence requirements.

---

Table of Contents

• Clause 4 — Context of the Organization
• Clause 5 — Leadership
• Clause 6 — Planning
• Clause 7 — Support
• Clause 8 — Operation
• Clause 9 — Performance Evaluation
• Clause 10 — Improvement
• Annex A Controls
• Cross-Standard Mapping
• EU AI Act Alignment

---

Clause 4 — Context of the Organization

4.1 Understanding the Organization and Its Context

Requirement: Determine external and internal issues relevant to the organization's purpose and that affect its ability to achieve the intended outcomes of its AIMS.

AI-Specific Issues to Document:

• External: Regulatory landscape (EU AI Act, national AI strategies), industry standards, public trust in AI, competitor AI capabilities, technology evolution, societal expectations for responsible AI
• Internal: AI maturity level, data infrastructure, talent availability, organizational culture toward AI adoption, existing management systems (ISO 27001, 9001)

Implementation Steps: 1. Conduct environmental scan for AI-related factors 2. Assess organizational AI maturity using a maturity model 3. Identify regulatory requirements applicable to AI systems 4. Document stakeholder expectations regarding AI 5. Review annually and update when context changes materially

Evidence Required:

• Context analysis document
• PESTLE or similar environmental analysis for AI factors
• AI maturity assessment results

Audit Questions:

• How has the organization identified external issues affecting its use of AI?
• What internal factors have been considered in establishing the AIMS?
• How does the organization monitor changes in its AI context?

Common Findings:

• Context analysis is generic and does not address AI-specific issues
• No consideration of societal impact or public trust factors
• Context not reviewed when new AI systems are deployed

4.2 Understanding the Needs and Expectations of Interested Parties

Requirement: Determine interested parties relevant to the AIMS and their requirements.

Key Interested Parties for AI:

Evidence Required:

• Interested party register with requirements
• Stakeholder engagement records
• Regulatory requirement mapping

Audit Questions:

• Who are the interested parties for each AI system?
• How are their requirements determined and monitored?
• How are conflicting requirements between parties resolved?

4.3 Determining the Scope of the AIMS

Requirement: Determine the boundaries and applicability of the AIMS to establish its scope.

Scope Considerations:

• Which AI systems are included (all, or subset by risk level)
• Organizational units involved in AI development, deployment, and use
• Third-party AI systems used by the organization
• Geographic boundaries and regulatory jurisdictions
• Integration with existing management systems (ISO 27001, ISO 9001)

Evidence Required:

• AIMS scope statement document
• AI system inventory within scope
• Justification for any exclusions

Audit Questions:

• What AI systems are included in the AIMS scope?
• Are there AI systems excluded from scope? What is the justification?
• Does the scope cover third-party AI components and services?

4.4 AI Management System

Requirement: Establish, implement, maintain, and continually improve the AIMS in accordance with the standard.

Evidence Required:

• AIMS manual or integrated documentation
• Process maps for AIMS
• Continual improvement records

---

Clause 5 — Leadership

5.1 Leadership and Commitment

Requirement: Top management shall demonstrate leadership and commitment with respect to the AIMS.

Demonstration Methods:

• Allocating adequate resources (budget, personnel, compute, data)
• Establishing AI policy and objectives
• Integrating AIMS requirements into business processes
• Promoting continual improvement and responsible AI
• Supporting other relevant management roles
• Communicating the importance of effective AI governance

Evidence Required:

• Management meeting minutes showing AI governance topics
• Resource allocation records (budget, headcount)
• Management communications on AI policy

Audit Questions:

• How does top management demonstrate commitment to the AIMS?
• What resources have been allocated for AI governance?
• How is AI governance integrated into business strategy?

Common Findings:

• Management commitment is stated but not substantiated with resource allocation
• AI governance is delegated without management oversight
• No regular management review of AIMS effectiveness

5.2 Policy

Requirement: Establish an AI policy that is appropriate, provides a framework for objectives, includes commitment to satisfy requirements, and includes commitment to continual improvement.

AI Policy Must Include:

• Commitment to responsible and ethical AI development and use
• Principles for fairness, transparency, safety, privacy, and accountability
• Alignment with organizational values and applicable regulations
• Framework for setting and reviewing AI objectives
• Commitment to continual improvement of the AIMS
• Commitment to satisfying applicable regulatory requirements
• Statement on human oversight of AI decisions

Evidence Required:

• Documented AI policy signed by top management
• Communication records (email, intranet, training)
• Policy review records

Audit Questions:

• Does the AI policy cover responsible AI principles?
• How is the policy communicated across the organization?
• When was the policy last reviewed and updated?
• Do employees understand and apply the policy?

5.3 Organizational Roles, Responsibilities, and Authorities

Requirement: Ensure responsibilities and authorities for relevant roles are assigned, communicated, and understood.

Key Roles:

Evidence Required:

• Organizational chart with AIMS roles
• Role descriptions with authorities
• Appointment records

---

Clause 6 — Planning

6.1 Actions to Address Risks and Opportunities

Requirement: Determine risks and opportunities relevant to the AIMS and plan actions to address them.

AI-Specific Risk Categories:

Risk Assessment Methodology Must Define:

• Risk identification process (how risks are discovered)
• Risk analysis (likelihood and impact assessment criteria)
• Risk evaluation (risk acceptance criteria and thresholds)
• Risk treatment options (Avoid, Mitigate, Transfer, Accept)
• Risk ownership and accountability
• Risk review frequency

Evidence Required:

• Risk assessment methodology document
• Risk register with AI-specific risks
• Risk treatment plans
• Risk acceptance records

6.2 AI Management System Objectives and Planning to Achieve Them

Requirement: Establish measurable AI objectives and plan how to achieve them.

Example AI Objectives:

Evidence Required:

• Documented AI objectives (SMART criteria)
• Plans to achieve objectives (who, what, when, resources)
• Monitoring and measurement records

---

Clause 7 — Support

7.1 Resources

AI-Specific Resources:

• Compute: GPU/TPU capacity, cloud resources, training infrastructure
• Data: Training datasets, evaluation datasets, production data pipelines
• Expertise: Data scientists, ML engineers, AI ethics specialists, domain experts
• Tools: ML platforms, monitoring tools, testing frameworks, version control
• Infrastructure: Development environments, staging, production, backup

7.2 Competence

Required Competencies:

Evidence Required:

• Competence requirements per role
• Training records and certifications
• Competence evaluation results

7.3 Awareness

All relevant persons must be aware of:

• The AI policy
• Their contribution to AIMS effectiveness
• Implications of not conforming to AIMS requirements
• Responsible AI principles and their practical application

7.5 Documented Information

Required Documented Information:

---

Clause 8 — Operation

8.1 Operational Planning and Control

Requirement: Plan, implement, and control processes needed to meet requirements and implement actions from Clause 6.

8.2 AI Risk Assessment

Requirement: Perform AI risk assessments at planned intervals or when significant changes are proposed.

When to Conduct Risk Assessment:

• Before developing a new AI system
• Before deploying an AI system to production
• When significant changes are made to an AI system
• After an AI incident
• At planned intervals (annually minimum)
• When the external context changes materially

8.3 AI Risk Treatment

Treatment Options:

8.4 AI System Lifecycle

See: ai-lifecycle-management.md for detailed lifecycle guidance.

Evidence Required:

• Lifecycle procedures per stage
• Design records (requirements, architecture, ethical review)
• Development records (data preparation, model training, code review)
• Testing records (functional, bias, robustness, performance)
• Deployment records (approval, staging, monitoring setup)
• Operational records (monitoring logs, performance reports)
• Retirement records (decommissioning plan, data disposal)

---

Clause 9 — Performance Evaluation

9.1 Monitoring, Measurement, Analysis, and Evaluation

AI Performance Metrics to Monitor:

9.2 Internal Audit

Audit Program Requirements:

• Planned intervals (annually minimum)
• Covers all AIMS clauses and applicable Annex A controls
• Auditor independence from area being audited
• Audit criteria, scope, frequency, and methods defined
• Results reported to relevant management

Audit Evidence Checklist:

• [ ] AIMS scope and policy documents
• [ ] Risk assessments and treatment plans
• [ ] AI system inventory and lifecycle records
• [ ] Competence and training records
• [ ] Monitoring and measurement results
• [ ] Incident and nonconformity records
• [ ] Management review minutes
• [ ] Statement of Applicability

9.3 Management Review

Inputs to Management Review:

• Status of actions from previous reviews
• Changes in internal/external issues
• AI system performance and trends
• Audit results
• Nonconformities and corrective actions
• Monitoring and measurement results
• AI incident reports
• Interested party feedback
• Opportunities for improvement

Outputs from Management Review:

• Decisions on improvement opportunities
• Changes to AIMS (scope, policy, objectives)
• Resource allocation decisions
• Risk appetite adjustments

---

Clause 10 — Improvement

10.1 Nonconformity and Corrective Action

Process: 1. Identify nonconformity 2. React to control and correct it 3. Evaluate need for action to eliminate cause 4. Implement corrective action 5. Review effectiveness 6. Update risks and opportunities if needed 7. Make changes to AIMS if needed

10.2 Continual Improvement

Improvement Sources:

• Internal audit findings
• Management review outputs
• AI incident lessons learned
• Industry best practice evolution
• Regulatory changes
• Technology advancement
• Stakeholder feedback

10.3 AI Incident Management (AI-Specific)

Incident Types:

• Model producing biased or discriminatory outputs
• Unexpected model behavior or hallucination
• Data breach involving AI training/inference data
• AI system causing harm (physical, financial, psychological)
• Adversarial attack on AI system
• AI system producing unsafe outputs
• Significant performance degradation

Incident Management Process: 1. Detection and reporting 2. Classification (severity, impact, scope) 3. Containment (disable, rollback, human override) 4. Investigation (root cause analysis) 5. Resolution and recovery 6. Communication to interested parties 7. Lessons learned and preventive actions 8. Documentation and evidence preservation

---

Annex A Controls

A.2 — AI Policies

Objective: Provide management direction and support for AI in accordance with business requirements and applicable laws.

Requirements:

• Define policies for AI development, deployment, and use
• Include responsible AI principles (fairness, transparency, safety, privacy)
• Approve policies at management level
• Communicate to all relevant persons
• Review at planned intervals

Evidence: AI policy document, communication records, review records

Audit Questions:

• Is there a documented AI policy?
• Does it cover responsible AI principles?
• Is it communicated to all employees involved with AI?
• When was it last reviewed?

A.3 — Internal Organization for AI

Objective: Establish a management framework to initiate and control AI implementation.

Requirements:

• Define roles and responsibilities for AI governance
• Implement segregation of duties (development vs. validation vs. deployment)
• Establish AI governance committee or board
• Define escalation and decision-making authority

Evidence: Org chart, role descriptions, governance charter, meeting minutes

A.4 — Resources for AI Systems

Objective: Achieve and maintain appropriate protection and management of resources.

Requirements:

• Identify and provide resources for each AI system
• Manage compute, data, and expertise resources
• Plan for resource scaling and capacity
• Address resource dependencies and single points of failure

Evidence: Resource inventory, capacity plans, budget allocation

A.5 — Assessing AI System Impact

Objective: Identify and evaluate impacts of AI systems on individuals and society.

Requirements:

• Establish impact assessment process
• Conduct assessments before deployment and periodically thereafter
• Evaluate fairness, safety, privacy, transparency, and societal impact
• Document assessment results and treatment decisions
• Review when systems change materially

Evidence: Impact assessment procedure, completed assessments, treatment records

A.6 — AI System Lifecycle

Objective: Ensure controls throughout the AI system lifecycle.

Requirements:

• Define lifecycle stages and gates
• Implement design, development, testing, deployment, monitoring, and retirement controls
• Version control for models and data
• Change management procedures
• Testing and validation requirements

Evidence: Lifecycle procedure, stage-gate records, version history, test results

A.7 — Data for AI Systems

Objective: Ensure appropriate management of data used in AI systems.

Requirements:

• Data quality assessment and improvement
• Data provenance and lineage tracking
• Bias assessment for training and evaluation data
• Data governance and access controls
• Personal data protection measures
• Data retention and disposal
• Data labeling quality assurance

Evidence: Data governance policy, quality reports, provenance records, bias assessments

A.8 — Information for Interested Parties

Objective: Provide appropriate transparency about AI systems.

Requirements:

• Disclose use of AI to affected individuals
• Provide information about AI system capabilities and limitations
• Enable meaningful human oversight
• Publish transparency reports or model cards
• Communicate AI-related incidents to affected parties

Evidence: Disclosure mechanisms, model cards, transparency reports, incident communications

A.9 — Use of AI Systems

Objective: Ensure responsible use of AI systems.

Requirements:

• Define acceptable use policies for AI systems
• Implement human oversight mechanisms
• Provide user guidance and training
• Monitor use for compliance with acceptable use
• Prevent misuse through technical and organizational controls

Evidence: Acceptable use policy, human oversight procedures, user guidance, monitoring records

A.10 — Third-Party Relationships

Objective: Manage risks associated with third-party AI components and services.

Requirements:

• Evaluate third-party AI suppliers before engagement
• Include AI-specific requirements in contracts
• Monitor third-party compliance with requirements
• Manage AI supply chain risks
• Conduct due diligence on third-party training data

Evidence: Supplier evaluation records, contracts, monitoring records, due diligence reports

---

Cross-Standard Mapping

ISO 42001 to ISO 27001 Mapping

ISO 42001 to ISO 9001 Mapping

---

EU AI Act Alignment

High-Risk AI Systems (Articles 6-15, 17)

Using ISO 42001 Certification for EU AI Act Compliance

ISO 42001 certification can support EU AI Act conformity assessment:

• Demonstrates a functioning quality management system (Art. 17)
• Provides evidence of risk management (Art. 9)
• Shows data governance practices (Art. 10)
• Documents human oversight procedures (Art. 14)

Note: ISO 42001 certification alone does not guarantee EU AI Act compliance. Additional technical requirements specific to the AI system must also be met.