OECD AI Principles
Internationally agreed principles from the OECD for how AI should be developed and used responsibly, covering fairness, transparency, accountability, and safety.
The Simple Version
Internationally agreed principles from the OECD for how AI should be developed and used responsibly, covering fairness, transparency, accountability, and safety.
Detailed Explanation
The OECD Recommendation on AI (adopted May 2019, updated 2024) provides five principles for responsible AI: inclusive growth and sustainable development; human-centred values and fairness; transparency and explainability; robustness, security, and safety; and accountability. The 2024 update expanded definitions and guidance to address generative AI and GPAI models. The EU AI Act's Article 3 definition of 'AI system' aligns with the OECD definition. The OECD AI Principles are the foundational international normative framework on which most subsequent AI regulation and standards are built.
Key Characteristics
- Five principles covering values, transparency, robustness, and accountability
- Adopted by 46 OECD and partner countries, the widest international AI policy consensus
- EU AI Act Article 3 AI system definition aligns with OECD definition
- Updated in 2024 to address generative AI and GPAI
Why It Matters
Organisations operating globally can use the OECD AI Principles as a values framework that bridges different national regulatory requirements, they underpin both the EU AI Act and the NIST AI RMF.
Real-World Analogy
Like the UN Guiding Principles on Business and Human Rights, an internationally agreed normative framework that shapes national regulation and corporate policy without being directly legally binding.
Common Misconceptions
- The OECD AI Principles are legally binding, they are a Recommendation, not a binding treaty; their influence is through national implementation and voluntary adoption.
- The Principles are mainly relevant to OECD member countries, they have been adopted by many non-member partner countries and inform global AI regulation design.