OECD Recommendation of the Council on Artificial Intelligence (OECD AI Principles)

What it is

The Recommendation of the Council on Artificial Intelligence, known as the OECD AI Principles, was adopted by the OECD Council in May 2019. It was the first intergovernmental standard on AI. It was revised in May 2024 to reflect developments in general-purpose and generative AI.

The Recommendation has two parts. The first sets out five values-based principles for the responsible stewardship of trustworthy AI. The second sets out five recommendations to governments on national policy and international cooperation. Adherence is open to non-member countries, and the principles also formed the basis of the G20 AI Principles adopted in 2019.

The OECD maintains the principles and related tools through OECD.AI, its policy observatory, including a catalogue of tools and metrics for trustworthy AI.

Who it applies to

Formally, the Recommendation is addressed to adherent governments. The five values-based principles are expressed as expectations of all AI actors, meaning anyone who plays an active role in the AI system lifecycle, including organisations that deploy or operate AI. No organisation is legally bound by them, but organisations meet them indirectly through laws and frameworks that build on them, and directly when they adopt them as the basis of their own AI policy.

What it requires

The five values-based principles

1. Inclusive growth, sustainable development and well-being. AI should benefit people and the planet, advancing inclusion, reducing inequality and protecting the environment.
2. Human rights and democratic values, including fairness and privacy. AI actors should respect the rule of law, human rights, democratic values and diversity throughout the lifecycle, with safeguards including the capacity for human agency and oversight. The 2024 revision added attention to misinformation and disinformation amplified by AI.
3. Transparency and explainability. AI actors should provide meaningful information about AI systems, so people know when they are interacting with AI, understand outcomes where feasible, and can challenge outcomes that affect them adversely.
4. Robustness, security and safety. AI systems should be robust, secure and safe throughout their lifecycle. The 2024 revision added that mechanisms should be in place to override, repair or decommission a system safely if it risks causing undue harm or behaves in undesired ways.
5. Accountability. AI actors should be accountable for the proper functioning of AI systems, with traceability of datasets, processes and decisions, and a systematic risk management approach applied at each phase of the lifecycle, including cooperation between different AI actors.

The five recommendations for governments

1. Investing in AI research and development.
2. Fostering an inclusive AI-enabling ecosystem.
3. Shaping an enabling interoperable governance and policy environment for AI.
4. Building human capacity and preparing for labour market transition.
5. International co-operation for trustworthy AI.

Putting them into practice

For an organisation, the principles translate into familiar governance work. That means a policy that commits to them, a register of AI systems so accountability can be assigned, risk and impact assessments that consider fairness, privacy and safety, transparency to people affected, human oversight and the ability to stop a system, and records that make decisions traceable.

Key dates

The Recommendation was adopted in May 2019. The OECD updated its definition of an AI system in November 2023, and the Council revised the Recommendation in May 2024. The updated definition describes an AI system as a machine-based system that infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments, with varying levels of autonomy and adaptiveness after deployment.

How it relates to other frameworks

The OECD principles are a common ancestor for much of the AI governance landscape. The EU AI Act’s definition of an AI system is aligned with the OECD’s. The NIST AI Risk Management Framework’s trustworthiness characteristics closely follow them. Many national AI ethics frameworks draw on them. The OECD also runs the reporting framework for the G7 Hiroshima Process Code of Conduct, which turns the principles into specific commitments for developers of advanced AI. UNESCO’s Recommendation on the Ethics of AI covers similar values from the UN system.

Because they are high-level, the principles are best used as the values layer of a governance programme, with a management system such as ISO/IEC 42001 or a risk framework such as the NIST AI RMF supplying the operational detail.