AI Governance Audit Checklist
Comprehensive governance maturity assessment for M&E and international development organizations. Evaluate your AI governance readiness across 8 key domains based on 20 global frameworks.
EvalCommunity AI Academy • Module 1, Lesson 4: AI Governance Frameworks
Your Progress
0
Completed
64
Total Items
0%
Compliance
Current Governance Maturity Level:
InitialAI Governance Audit Checklist
How to Use This Audit Checklist
This comprehensive audit tool helps you assess your organization's AI governance maturity across 8 critical domains aligned with international frameworks including UNESCO, OECD, UNDP, and NIST.
- Review each category and check items that your organization has fully implemented
- Provide evidence in the text boxes to document your governance practices
- Be honest in your assessment - gaps are opportunities for improvement
- Generate your report to receive personalized recommendations and action plans
- Export your results to share with stakeholders or track progress over time
1. Global Normative & Policy Alignment
0/8
Assessment of alignment with global AI ethics frameworks and multilateral governance standards for international development and M&E contexts.
UNESCO AI Ethics Principles Adoption
Organization has formally adopted or mapped governance practices to UNESCO's Recommendation on the Ethics of AI, including human rights, transparency, and sustainability principles.
UNESCO
OECD AI Principles Compliance
AI systems and governance processes align with OECD AI Principles, particularly inclusive growth, human-centered values, and transparency requirements.
OECD
Human Rights Impact Assessment
Organization conducts human rights impact assessments for AI systems, particularly those affecting vulnerable populations or in humanitarian contexts.
UNESCO
UNDP
SDG Alignment Documentation
AI initiatives are explicitly mapped to Sustainable Development Goals with documented contribution pathways and impact metrics.
UNDP
UN Global Compact
Donor Compliance Framework
AI governance policies align with major donor requirements (e.g., USAID, DFID, EU) and international development standards.
OECD-DAC
Environmental Sustainability Considerations
Organization assesses and manages the environmental impact of AI systems, including energy consumption and carbon footprint.
UNESCO
Multi-stakeholder Governance Structure
AI governance includes representation from diverse stakeholders including beneficiaries, technical staff, leadership, and external advisors.
UNDP
OECD
Cross-border Data Governance
Clear policies exist for cross-border data transfers and AI deployment in multiple jurisdictions, compliant with international standards.
OECD
2. Development & Humanitarian Context
0/8
Specific governance considerations for AI in development programming, humanitarian response, and vulnerable population protection.
"Leaving No One Behind" Assessment
AI systems are evaluated for their impact on marginalized and vulnerable populations, with mechanisms to prevent digital exclusion.
UNDP
Do-No-Harm Principle Implementation
Humanitarian AI projects follow do-no-harm principles with risk assessments for conflict sensitivity and unintended consequences.
ICRC
IFRC
Child Rights & Protection Standards
AI systems involving or affecting children comply with UNICEF policy guidance and UN Convention on the Rights of the Child.
UNICEF
Informed Consent Mechanisms
Clear, culturally appropriate informed consent processes exist for data collection and AI use, especially with vulnerable populations.
ICRC
UNESCO
Poverty Reduction Focus
AI initiatives explicitly target poverty reduction with documented theory of change and impact pathways for development outcomes.
World Bank
UNDP
Local Capacity Building Strategy
AI projects include plans for local capacity development and knowledge transfer to ensure sustainability and ownership.
UNDP
World Bank
Humanitarian Data Protection
Extra safeguards protect sensitive humanitarian data including data minimization, purpose limitation, and secure storage protocols.
ICRC
IFRC
Contextual Adaptation Framework
AI governance frameworks are adapted to local contexts, cultural norms, and institutional capacities rather than applied uniformly.
UNDP
3. M&E & Evaluation Standards
0/8
Application of evaluation principles and ethical standards to AI systems in M&E contexts.
OECD-DAC Criteria Applied to AI
AI systems are evaluated using OECD-DAC criteria (relevance, coherence, effectiveness, efficiency, impact, sustainability).
OECD-DAC
UNEG Ethical Guidelines Compliance
AI-enabled evaluations follow UNEG ethical guidelines including respect for people, beneficence, justice, and independence.
UNEG
Independence & Impartiality Safeguards
Clear separation exists between AI system development and evaluation, with independent oversight and conflict of interest management.
UNEG
Impact Measurement Framework
Systematic framework measures AI's contribution to development outcomes with baseline, targets, and attribution methodologies.
OECD-DAC
Stakeholder Participation in Evaluation
Beneficiaries and affected communities participate meaningfully in AI evaluation design, data collection, and interpretation.
UNEG
UNDP
Evaluation Quality Assurance
Quality assurance processes ensure AI evaluations meet professional standards with peer review and validation mechanisms.
UNEG
Equity & Fairness Evaluation
Systematic evaluation of AI systems' differential impacts across gender, age, disability, ethnicity, and socioeconomic groups.
OECD-DAC
UNDP
Learning & Adaptive Management
Evaluation findings feed into AI system improvements with documented learning loops and adaptive management processes.
OECD-DAC
4. Technical Standards & Risk Management
0/8
Technical governance frameworks, risk assessment methodologies, and management system standards for AI.
AI Risk Management Framework Implementation
Organization uses NIST AI RMF or equivalent lifecycle-based framework to govern, map, measure, and manage AI risks.
NIST AI RMF
ISO 42001 Compliance or Certification
AI management system follows ISO/IEC 42001 standards with documented Plan-Do-Check-Act cycle and continual improvement.
ISO 42001
AI Impact Assessment Process
Comprehensive AI Impact Assessments (AIIA) conducted before deployment covering technical, ethical, and social dimensions.
NIST
UNESCO
Model Validation & Testing Procedures
Systematic testing including red-team testing, robustness checks, and adversarial testing before AI deployment.
NIST
Bias Detection & Mitigation
Technical processes detect and mitigate bias using fairness metrics, demographic parity assessments, and bias audits.
NIST
UNESCO
AI Risk Register Maintenance
Active AI risk register documents identified risks, mitigation strategies, ownership, and regular review cycles.
NIST
ISO 42001
Security & Cybersecurity Controls
AI systems protected by robust cybersecurity controls including data encryption, access management, and threat monitoring.
NIST
ISO 42001
Performance Monitoring & Drift Detection
Continuous monitoring detects model drift, performance degradation, and unexpected behaviors with automated alerts.
NIST
ISO 42001
5. Transparency & Accountability
0/8
Documentation, explainability, and accountability mechanisms for AI systems and decision-making processes.
Model Cards & Documentation
All AI models documented with model cards including intended use, limitations, performance metrics, and ethical considerations.
Model Cards
Explainability & Interpretability
AI decisions can be explained to affected stakeholders in understandable language, with mechanisms for questioning outcomes.
OECD
UNESCO
RACI Matrix for AI Governance
Clear RACI (Responsible, Accountable, Consulted, Informed) matrix defines roles and accountabilities for AI decisions.
RACI
Public Transparency Reporting
Organization publishes regular transparency reports on AI use, incidents, and governance performance for stakeholders.
OECD
UNESCO
Audit Trails & Logging
Comprehensive audit trails track AI decisions, data usage, model changes, and governance actions with tamper-proof logging.
ISO 42001
Grievance & Redress Mechanisms
Accessible mechanisms allow stakeholders to challenge AI decisions, report concerns, and seek redress for harms.
UNESCO
OECD
Algorithmic Decision Documentation
All automated or AI-assisted decisions documented with rationale, data sources, and human oversight involvement.
OECD
External Audit & Third-Party Review
Independent external audits or third-party reviews assess AI governance compliance and effectiveness periodically.
ISO 42001
6. Data Governance & Privacy
0/8
Data protection, privacy safeguards, and responsible data management practices for AI systems.
Data Protection Impact Assessment (DPIA)
DPIAs conducted for all AI systems processing personal data, identifying and mitigating privacy risks before deployment.
DPIA
ICRC
Data Minimization Principle
Only minimum necessary data collected and processed for specified AI purposes, with regular data deletion/archiving.
ICRC
GDPR
Purpose Limitation & Use Restrictions
Data used only for explicitly stated purposes with controls preventing secondary use or function creep.
ICRC
UNESCO
Data Quality & Accuracy Standards
Processes ensure training and operational data are accurate, complete, representative, and regularly validated.
NIST
ISO 42001
Sensitive Data Protections
Enhanced safeguards protect sensitive categories (health, children, conflict-affected) with encryption and access controls.
UNICEF
ICRC
Data Subject Rights Implementation
Mechanisms enable individuals to access, correct, delete, or port their data used in AI systems.
GDPR
UNESCO
Third-Party Data Sharing Governance
Clear policies govern data sharing with third parties including contracts, security requirements, and accountability measures.
ICRC
OECD
Data Breach Response Plan
Documented incident response plan for data breaches including detection, containment, notification, and remediation procedures.
ICRC
ISO 42001
7. Organizational Capacity & Culture
0/8
Human resources, training, capacity building, and organizational culture for responsible AI.
AI Ethics Training Program
Mandatory AI ethics training for all staff working with or affected by AI systems, with regular refresher courses.
UNESCO
UNDP
Technical Capacity Development
Ongoing technical training builds staff capacity in AI literacy, data science, and governance implementation.
UNDP
World Bank
Dedicated AI Governance Roles
Clear roles and resources allocated for AI governance including ethics officers, data protection officers, or governance leads.
ISO 42001
RACI
Ethics Committee or Advisory Board
Active AI ethics committee with diverse membership provides guidance on ethical dilemmas and governance decisions.
UNESCO
UNDP
Diverse & Inclusive AI Teams
AI development and governance teams reflect diversity in gender, age, ethnicity, discipline, and lived experience.
UNESCO
OECD
Knowledge Management Systems
Systems capture and share AI governance lessons, best practices, and institutional knowledge across the organization.
ISO 42001
UNDP
Responsible AI Culture Indicators
Organizational culture values and rewards ethical AI practices with mechanisms to raise concerns without retaliation.
UNESCO
OECD
External Expertise & Partnerships
Organization accesses external AI ethics and governance expertise through partnerships, advisors, or consultants.
UNDP
8. Continuous Improvement & Learning
0/8
Systems and processes for ongoing governance improvement, incident learning, and adaptive management.
Regular Governance Reviews
Systematic reviews of AI governance effectiveness conducted quarterly or bi-annually with documented improvements.
ISO 42001
OECD-DAC
Incident Reporting & Learning System
All AI incidents (errors, bias, harms) tracked with root cause analysis and lessons incorporated into governance.
NIST
ISO 42001
Stakeholder Feedback Mechanisms
Regular feedback collected from AI users, beneficiaries, and affected communities to improve governance practices.
UNEG
UNDP
Emerging Risk Monitoring
Active scanning for emerging AI risks, technologies, and governance best practices with timely policy updates.
NIST
OECD
Governance Metrics Dashboard
Real-time dashboard tracks key governance metrics including compliance rates, incident frequency, and training completion.
ISO 42001
Policy Version Control & Updates
AI governance policies maintain version control with regular updates reflecting new learnings, technologies, and standards.
ISO 42001
Benchmarking & Peer Learning
Organization participates in peer learning networks and benchmarks governance practices against sector leaders.
OECD
UNDP
Innovation & Experimentation Framework
Structured approach to testing new governance approaches with sandboxes, pilots, and controlled experimentation.
OECD