1. Inclusive growth, sustainable development, well-being
AI should contribute to inclusive growth, sustainable development, well-being for people and planet:
- Inclusive growth: AI should reduce inequalities, not amplify.
- Sustainable development: AI should support SDGs.
- Well-being: AI should advance human well-being broadly defined.
2024 update emphasized sustainability dimension.
2. Human-centred values and fairness
AI should respect:
- Rule of law
- Human rights
- Democratic values
Including (non-exhaustive):
- Freedom
- Dignity
- Autonomy
- Privacy and data protection
- Non-discrimination and equality
- Diversity
- Fairness
- Social justice
- Internationally recognised labour rights
Mechanisms include human oversight, particularly for high-stakes AI.
3. Transparency and explainability
AI should be transparent and explainable:
- Information about AI systems: those affected should know AI is involved.
- Explainability: outcomes should be understandable.
- Documentation: AI systems' purposes, capabilities, limitations documented.
- Information for affected parties: appropriate to outcomes and stakes.
4. Robustness, security and safety
AI should be robust, secure, safe throughout lifecycle:
- Reliability: consistent operation as intended.
- Security: resistance to adversarial attack.
- Safety: no unreasonable risk.
- Traceability: ability to audit decisions.
2024 update strengthened with information integrity (countering mis/disinformation).
5. Accountability
AI actors should be accountable:
- Responsibility for proper functioning.
- Risk management throughout lifecycle.
- Adherence to other principles.
- Cooperation with regulators.
- Pathways for affected parties to seek redress.
Connection to operational frameworks
Each principle implemented operationally via:
- Inclusive growth / sustainable development: ISO 42001 impact assessment Annex A.5; NIST AI RMF GenAI Profile Category 5 (environmental) + Category 6 (bias).
- Human-centred values: ISO 42001 A.5; EU AI Act Article 26 (deployer obligations) + Annex IV; GDPR.
- Transparency: ISO 42001 A.8; EU AI Act Article 50 + Article 13; NIST AI RMF.
- Robustness, security, safety: OWASP LLM Top 10; MITRE ATLAS; ISO 42001 A.6; EU AI Act Article 15.
- Accountability: ISO 42001 Cl 5; NIST AI RMF GOVERN function; ISO 27001 Cl 5.
SRE and AI-agent fit notes
For AI feature design and operation:
- OECD principles as high-level checklist.
- Operational depth via NIST AI RMF / ISO 42001 / OWASP LLM Top 10.
Stefan-context implementation sketch
- For client AI work: OECD principles + operational frameworks combined.
- Vendor evaluation: OECD principles adherence as one signal.