Responsible and trustworthy AI principles

Every major framework converges on roughly the same principles. Learn the common core, then the NIST list exactly.

The common core

• Fairness and non-discrimination
• Safety
• Security and resilience
• Transparency and explainability
• Accountability
• Privacy
• Human oversight and agency
• Validity, reliability and robustness

NIST AI RMF: seven characteristics of trustworthy AI

1. Valid and reliable (the base the others sit on)
2. Safe
3. Secure and resilient
4. Accountable and transparent
5. Explainable and interpretable
6. Privacy-enhanced
7. Fair, with harmful bias managed

OECD AI Principles (2019, updated 2024)

Five values-based principles: inclusive growth and well-being; human rights and democratic values including fairness and privacy; transparency and explainability; robustness, security and safety; and accountability. Plus five recommendations to governments (research investment, an inclusive ecosystem, an enabling policy environment, human capacity, international cooperation).

Principles conflict

Accuracy can trade off against explainability; privacy techniques can reduce accuracy; fairness metrics can conflict with each other. Governance does not make these trade-offs vanish. It makes sure the right people decide them deliberately and document why.