The OECD AI Principles as the AIGP tests it

The OECD's Recommendation on Artificial Intelligence was adopted in 2019, four years before the NIST framework, and it is the source among the three that the AIGP's competency II.D names whose words turn up in the EU AI Act. Its definition of an AI system is the one the Act adopted; its five principles echo through the Body of Knowledge's own list of responsible-AI principles; and it is not law. The Body of Knowledge asks you to "Understand the Organisation for Economic Co-operation and Development (OECD) principles, framework, policies and recommended practices for trustworthy AI". This guide sets out what the Recommendation is, its two sections, what changed in 2024, and the handful of distinctions an exam question can turn on, all from the instrument's own text.

The short answer

• A Recommendation, not law: adopted May 22, 2019 and revised in 2023 and 2024; the OECD says its Recommendations "are not legally binding".
• Five principles for AI actors: inclusive growth and well-being; human rights and democratic values, including fairness and privacy; transparency and explainability; robustness, security and safety; accountability.
• Five recommendations for governments: research investment, an enabling ecosystem, an agile policy environment, human capacity, international cooperation.
• One definition, shared with the EU AI Act: a machine-based system that infers from its input how to generate outputs that can influence physical or virtual environments.
• The 2024 revision added information integrity, misuse, safety overrides and environmental sustainability, and moved traceability and risk management under Accountability.

What the Recommendation is

The Recommendation was adopted by the OECD Council on May 22, 2019, "on the proposal of the Digital Policy Committee", and its purpose is stated in one sentence: it "aims to foster innovation and trust in AI by promoting the responsible stewardship of trustworthy AI while ensuring respect for human rights and democratic values". A month later, "at the Osaka Summit, G20 Leaders welcomed the G20 AI Principles, drawn from the Recommendation". It has been revised twice: on November 8, 2023, "to update its definition of an 'AI System', in order to ensure the Recommendation continues to be technically accurate and reflect technological developments, including with respect to generative AI", and on May 3, 2024, by the Council meeting at ministerial level, "to reflect technological and policy developments, including with respect to generative AI, and to further facilitate its implementation".

Its legal weight is the first thing to know. The OECD's own note on its instruments says: "Recommendations are adopted by Council and are not legally binding. They represent a political commitment to the principles they contain and entail an expectation that Adherents will do their best to implement them." The Recommendation addresses "Members and non-Members adhering to this Recommendation (hereafter the 'Adherents')". A country that adheres makes a political commitment; a company that follows the principles complies with nothing, because there is nothing to comply with. On the exam, that is the line between the OECD principles and the EU AI Act.

The definitions the exam borrows

The Recommendation defines three terms the Body of Knowledge uses throughout.

An AI system "is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment." Set that beside Article 3(1) of the EU AI Act, "a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments", and the shared core is plain: machine-based, inference from input, outputs that influence environments, varying autonomy, possible adaptiveness. A question on the OECD's definition and on the EU AI Act's is one question, not two.

The AI system lifecycle "typically involves several phases that include to: plan and design; collect and process data; build model(s) and/or adapt existing model(s) to specific tasks; test, evaluate, verify and validate; make available for use/deploy; operate and monitor; and retire/decommission", and the phases "often take place in an iterative manner and are not necessarily sequential". That lifecycle is the spine of the AIGP's Domains III and IV.

AI actors "are those who play an active role in the AI system lifecycle, including organisations and individuals that deploy or operate AI". The principles are addressed to AI actors; the recommendations are addressed to governments.

Section 1: the five principles

The Recommendation "RECOMMENDS that Members and non-Members adhering to this Recommendation (hereafter the 'Adherents') promote and implement the following principles for responsible stewardship of trustworthy AI", and lists five.

1.1 Inclusive growth, sustainable development and well-being. Stakeholders "should proactively engage in responsible stewardship of trustworthy AI in pursuit of beneficial outcomes for people and the planet", and the text names them: "augmenting human capabilities and enhancing creativity, advancing inclusion of underrepresented populations, reducing economic, social, gender and other inequalities, and protecting natural environments".

1.2 Respect for the rule of law, human rights and democratic values, including fairness and privacy. AI actors "should respect the rule of law, human rights, democratic and human-centred values throughout the AI system lifecycle", which "include non-discrimination and equality, freedom, dignity, autonomy of individuals, privacy and data protection, diversity, fairness, social justice, and internationally recognised labour rights", and, since 2024, "addressing misinformation and disinformation amplified by AI". The mechanisms the principle asks for include "capacity for human agency and oversight".

1.3 Transparency and explainability. AI actors "should commit to transparency and responsible disclosure regarding AI systems", providing meaningful information to four ends: "to foster a general understanding of AI systems, including their capabilities and limitations"; "to make stakeholders aware of their interactions with AI systems, including in the workplace"; where feasible, to explain "the sources of data/input, factors, processes and/or logic that led to the prediction, content, recommendation or decision"; and "to provide information that enable those adversely affected by an AI system to challenge its output".

1.4 Robustness, security and safety. Systems "should be robust, secure and safe throughout their entire lifecycle so that, in conditions of normal use, foreseeable use or misuse, or other adverse conditions, they function appropriately and do not pose unreasonable safety and/or security risks". Since 2024 the principle adds that if systems "risk causing undue harm or exhibit undesired behaviour, they can be overridden, repaired, and/or decommissioned safely as needed", and that mechanisms should "bolster information integrity while ensuring respect for freedom of expression".

1.5 Accountability. AI actors "should be accountable for the proper functioning of AI systems and for the respect of the above principles, based on their roles, the context, and consistent with the state of the art". The principle now carries two things that used to sit elsewhere: traceability, "including in relation to datasets, processes and decisions made during the AI system lifecycle, to enable analysis of the AI system's outputs and responses to inquiry", and "a systematic risk management approach to each phase of the AI system lifecycle on an ongoing basis", with responsible business conduct and cooperation along the supply chain. The risks it names "include those related to harmful bias, human rights including safety, security, and privacy, as well as labour and intellectual property rights".

Section 2: the five recommendations to governments

The second section "RECOMMENDS that Adherents implement the following recommendations", and these are addressed to governments rather than AI actors:

• 2.1 Investing in AI research and development: "long-term public investment, and encourage private investment, in research and development and open science".
• 2.2 Fostering an inclusive AI-enabling ecosystem: "an inclusive, dynamic, sustainable, and interoperable digital ecosystem for trustworthy AI", including "data, AI technologies, computational and connectivity infrastructure, and mechanisms for sharing AI knowledge", with "data trusts" as an example of a sharing mechanism.
• 2.3 Shaping an enabling interoperable governance and policy environment for AI: "an agile policy environment that supports transitioning from the research and development stage to the deployment and operation stage", including experimentation in controlled settings.
• 2.4 "Building human capacity and preparing for labour market transformation": governments "should work closely with stakeholders to prepare for the transformation of the world of work and of society" and "empower people to effectively use and interact with AI systems".
• 2.5 International co-operation for trustworthy AI: governments "should actively co-operate to advance these principles and to progress on responsible stewardship of trustworthy AI".

The indicator names "principles, framework, policies and recommended practices", so both sections are in scope: the principles are what an organization adopts, and the recommendations are what governments do. A question can ask what the Recommendation's second section is, and the answer is national policy and international cooperation, not duties on companies.

What changed in 2024

The 2024 revision is recent enough to be worth knowing as a list, because a question written after the revision can use the revised text. The instrument's own summary says the updates aimed at "reflecting the growing importance of addressing misinformation and disinformation, and safeguarding information integrity in the context of generative AI"; "addressing uses outside of intended purpose, intentional misuse, or unintentional misuse"; "clarifying the information AI actors should provide regarding AI systems to ensure transparency and responsible disclosure"; "addressing safety concerns", so that systems "can be overridden, repaired, and/or decommissioned safely by human interaction"; "emphasising responsible business conduct throughout the AI system lifecycle"; "underscoring the need for jurisdictions to work together to promote interoperable governance and policy environments for AI"; and "introducing an explicit reference to environmental sustainability". It also moved "the text on traceability and risk management" into the Accountability principle "as the most appropriate principle for these concepts", which is the change the sample question above turns on.

How it is tested

The Body of Knowledge's indicator begins "Understand", so the questions are recognition and distinction, in four shapes:

1. Name the principle. A practice or a value is described; the answer is one of the five. The 2024 placements matter: information integrity sits under human rights and under safety, traceability and risk management under accountability.
2. The definition. The AI system definition, shared with the EU AI Act, and the lifecycle's phases.
3. Binding or not. The Recommendation is "not legally binding"; adherence is "a political commitment". Distractors describe fines, audits or certification, none of which the OECD has.
4. Who is addressed. Principles to AI actors, recommendations to governments; the G20 AI Principles were "drawn from the Recommendation".

The NIST AI RMF and the ISO standards are the other two sources in II.D; The NIST AI RMF as the AIGP tests it covers the framework, and The EU AI Act as the AIGP tests it the law that borrowed the definition. Our course teaches the Recommendation in its own lesson with practice questions in the exam's format; the first competency is free in full with a free account.