AI Ethical Certification (AIEC) – Assessor Checklist
Issued by IFPASA – a Global Initiative of the American Research Foundation
1. Transparency
- Model explainability and use of XAI techniques
- Availability of model documentation (e.g., datasheets, model cards)
- Disclosure of AI involvement in decisions
- Transparency in system limitations and scope
2. Accountability
- Clear assignment of responsibility
- Audit trail for AI decisions
- Grievance mechanisms in place
- Human oversight is enabled
3. Fairness & Bias Mitigation
- Bias detection and correction protocols
- Inclusive and representative training data
- Impact assessments across key demographics
- Ongoing fairness audits implemented
4. Privacy & Data Governance
- Data minimization and anonymization measures
- Consent and privacy controls are functional
- Compliance with GDPR/HIPAA/local laws
- Secure data handling and storage practices
5. Human-Centricity
- AI supports human agency and decision-making
- Design considers human dignity and autonomy
- Stakeholders engaged in design process
- System is accessible and user-friendly
6. Robustness & Safety
- System tested for reliability and stress conditions
- Detection of adversarial attacks/errors in place
- Fail-safe and fallback protocols exist
- Cybersecurity and threat monitoring active
7. Environmental Impact
- Efficient and sustainable model architecture
- Use of green infrastructure or resources
- Hardware lifecycle and resource consumption tracked
- Reporting on environmental metrics (optional for small entities)