OECD AI Principles

Overview

The Recommendation of the Council on Artificial Intelligence, often referred to as the OECD AI Principles, represents a landmark intergovernmental standard for the responsible development and deployment of artificial intelligence. Adopted by the OECD Council on May 22, 2019, this document was the first of its kind, establishing a global consensus on how to foster innovation and trust in AI while upholding human rights and democratic values. Its primary purpose is to guide policymakers and AI actors in navigating the complex ethical and societal implications of AI technologies. The Recommendation provides a foundational framework that balances the immense potential of AI for inclusive growth, sustainable development, and well-being with the imperative to mitigate risks such as bias, discrimination, privacy infringements, and security vulnerabilities.

Since its initial adoption, the Recommendation has undergone significant revisions to ensure its continued relevance in a rapidly evolving technological landscape. Notably, the OECD Council revised the document on November 8, 2023, to update the definition of an “AI System,” specifically to encompass generative AI and other technological advancements. Further comprehensive revisions were adopted on May 3, 2024, reflecting broader technological and policy developments and aiming to facilitate its implementation. These updates underscore the OECD's commitment to maintaining a robust and adaptable standard that can address emerging challenges and opportunities presented by AI, ensuring that the principles remain fit for purpose in guiding the global AI governance discourse. The Recommendation serves as a crucial blueprint, influencing national AI strategies and international legislative efforts worldwide.

Definitions

The Recommendation of the Council on Artificial Intelligence establishes a common understanding of several key terms crucial for its application and interpretation. Central to the document is the definition of an “AI System,” which was notably revised in November 2023 and May 2024 to ensure technical accuracy and reflect advancements, including generative AI. An AI System is defined as 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. This definition also acknowledges that different AI systems possess varying levels of autonomy and adaptiveness post-deployment. This updated understanding is vital for ensuring the Recommendation's applicability across the diverse and rapidly expanding spectrum of AI technologies.

Beyond the core definition of an AI System, the Recommendation also proposes a common understanding of the “AI System Lifecycle” and “AI Actors.” The AI System Lifecycle encompasses the entire spectrum of activities and processes relevant to an AI system, from its initial conception and design through to its deployment, operation, and eventual decommissioning. This comprehensive scope includes critical stages such as data collection, processing, model development, testing, validation, deployment, monitoring, and maintenance. “AI Actors” refer to the various individuals, organisations, or other entities that play a role at any point within this AI system lifecycle, including, but not limited to, developers, deployers, operators, and users of AI systems. These definitions are fundamental to delineating responsibilities and ensuring consistent application of the principles across all stakeholders involved in the AI ecosystem.

Governance and Institutional Framework

The governance and institutional framework supporting the OECD Recommendation on Artificial Intelligence is multifaceted and designed to promote its effective implementation and ongoing relevance. The Recommendation itself was adopted by the OECD Council on the proposal of the Digital Policy Committee (DPC), formerly known as the Committee on Digital Economy Policy (CDEP). This committee plays a central role in overseeing the OECD's work on digital policy, including AI. To further support the Recommendation's implementation, the Council instructed the DPC to develop practical guidance, establish a forum for information exchange on AI policy, and foster multi-stakeholder and interdisciplinary dialogue. This directive led to the creation of key institutional mechanisms that facilitate global cooperation and knowledge sharing in AI governance.

A significant component of this framework is the OECD.AI Policy Observatory, launched in February 2020. The OECD.AI serves as an inclusive hub for public policy on AI, providing multidisciplinary, evidence-based policy analysis by combining resources from across the OECD and its partners. It features a live database of national AI strategies, policies, and initiatives, enabling interactive comparison and continuous updates with AI metrics and good practices. Complementing the Observatory is the informal OECD Network of Experts on AI (ONE AI), an expert group comprising representatives from government, business, academia, and civil society, which offers AI-specific policy expertise and facilitates in-depth discussions. In 2022, recognising the expanding scope of its AI work, the DPC established the dedicated Working Party on AI Governance (AIGO), which specifically supports the Recommendation's implementation, develops practical guidance, and fosters multi-stakeholder dialogue through the OECD.AI Network of Experts. This robust institutional structure ensures continuous engagement and adaptation to the dynamic field of AI governance.

Key Provisions

The OECD Recommendation on Artificial Intelligence is structured around two substantive sections: five high-level, values-based principles for the responsible stewardship of trustworthy AI, and five recommendations for national policies and international co-operation. The values-based principles are designed to be universally applicable to all stakeholders throughout the AI system lifecycle. These principles include: i) inclusive growth, sustainable development, and well-being, advocating for AI to contribute positively to people and the planet; ii) respect for the rule of law, human rights, and democratic values, including fairness and privacy, emphasising non-discrimination, equality, and individual autonomy; iii) transparency and explainability, calling for meaningful information disclosure to foster understanding and challenge AI outcomes; iv) robustness, security, and safety, requiring AI systems to function reliably and safely without posing unreasonable risks; and v) accountability, mandating clear responsibilities for AI actors for the proper functioning and impact of AI systems. These principles collectively form the ethical bedrock for trustworthy AI.

Complementing these principles are five recommendations for national policies and international co-operation, which Adherents are encouraged to implement. These recommendations aim to create an enabling environment for trustworthy AI. They include: i) investing in AI research and development, promoting both public and private investment in technical and ethical AI research; ii) fostering an inclusive AI-enabling ecosystem, which involves developing access to digital technologies, infrastructure, and mechanisms for sharing AI knowledge; iii) shaping an enabling interoperable governance and policy environment for AI, supporting the deployment of trustworthy AI systems; iv) building human capacity and preparing for labour market transformation, addressing the impact of AI on employment and ensuring a skilled workforce; and v) international co-operation for trustworthy AI, encouraging collaborative efforts among governments and stakeholders to advance these principles globally and promote consensus-driven technical standards. These provisions collectively offer a comprehensive framework for guiding AI development and deployment towards beneficial societal outcomes.

Scope and Application

The Recommendation of the Council on Artificial Intelligence is designed with a broad and inclusive scope, aiming to be applicable across diverse contexts and stakeholders involved in the AI ecosystem. It focuses specifically on policy issues unique to AI, while also complementing existing OECD standards related to privacy, data protection, digital security risk management, and responsible business conduct. The Recommendation strives to set a standard that is both implementable and flexible, allowing it to remain relevant in the rapidly evolving field of AI. Its principles for responsible stewardship of trustworthy AI are intended for all stakeholders, including governments, businesses, civil society, and researchers, recognising their respective roles throughout the AI system lifecycle.

Geographically, the Recommendation extends its influence beyond OECD Member countries. While initially adopted by the OECD Council, it explicitly encourages adherence by non-Members, fostering a wider international alignment on AI governance. As of 2024, over 46 countries, including all G7 and G20 nations, have formally adhered to or endorsed the Principles, signifying its broad international acceptance and impact. This widespread adoption means that the Recommendation serves as a foundational blueprint for national AI strategies and has influenced major legislative efforts globally, such as the EU AI Act. The general nature of the principles ensures their applicability to AI developments worldwide, promoting a human-centred approach to AI that fosters research, preserves economic incentives for innovation, and applies consistently across various sectors and societies.

Implementation Framework

The implementation framework for the OECD Recommendation on Artificial Intelligence is designed to translate its high-level principles into practical, actionable guidance for governments and AI actors. Following the Recommendation's adoption, the OECD Council specifically instructed the Digital Policy Committee (DPC) to develop practical guidance, establish a forum for information exchange, and foster multi-stakeholder dialogue to support its implementation. This led to the launch of the OECD.AI Policy Observatory in February 2020, which serves as a central hub for sharing AI knowledge, policies, and best practices among countries and stakeholders. The Observatory is continuously updated with AI metrics, measurements, policies, and good practices, which in turn inform further practical guidance for implementation.

Further strengthening the implementation framework, the OECD created the Working Party on AI Governance (AIGO) in 2022. AIGO's mandate includes directly supporting the implementation of the Recommendation, developing practical guidance, and facilitating multi-stakeholder and interdisciplinary dialogue through the OECD.AI Network of Experts. Beyond these institutional mechanisms, the AIGN AI Governance Framework has emerged as a tool that operationalises the OECD AI Principles. AIGN is built as a full-stack, operational governance system that translates the OECD's soft-law principles into auditable, certifiable, and sector-ready practices, thereby bridging the gap between aspirational ethics and measurable implementation. This framework provides concrete modules and capabilities, such as ethics-by-design toolkits, model resilience testing, logging templates, and accountability matrices, to help organisations embed the principles into their AI development and deployment processes. The emphasis on practical guidance and operational tools underscores the OECD's commitment to ensuring that the Recommendation's principles are not just theoretical but can be effectively integrated into real-world AI governance practices.

Monitoring and Evaluation

The OECD Recommendation on Artificial Intelligence incorporates mechanisms for ongoing monitoring and evaluation to ensure its continued relevance and effectiveness in the face of rapid technological advancements. The dynamic nature of AI necessitates a flexible policy environment that can adapt to new developments, and the OECD has demonstrated this commitment through periodic reviews and revisions of the Recommendation. A key aspect of this monitoring process is the regular reporting to the OECD Council on the Recommendation's implementation, dissemination, and continued relevance. These reports serve as a basis for assessing how well the principles are being applied by Adherents and identifying areas where updates or further guidance may be needed.

Evidence of this continuous monitoring and evaluation is seen in the revisions made to the Recommendation since its initial adoption. For instance, the definition of an “AI System” was updated on November 8, 2023, specifically to ensure it remained technically accurate and reflected technological developments, including the emergence of generative AI. Furthermore, based on the findings and conclusions of the 2024 Report to Council, the Recommendation underwent more extensive revisions on May 3, 2024, to incorporate broader technological and policy developments and to further facilitate its implementation. This iterative process of review, reporting, and revision, supported by the work of the OECD.AI Policy Observatory and its expert networks, ensures that the Recommendation remains a robust and timely intergovernmental standard for trustworthy AI governance.

Relationship to Other Instruments

The OECD Recommendation on Artificial Intelligence does not exist in isolation but rather complements and builds upon a broader landscape of international legal and policy instruments. The document explicitly states that it complements existing OECD standards already relevant to AI, such as those concerning privacy and data protection, digital security risk management, and responsible business conduct. This approach ensures coherence across the OECD's normative framework, integrating AI-specific considerations within established principles for digital and economic governance. For example, the Recommendation references the OECD Guidelines for Multinational Enterprises, the Recommendation of the Council concerning Guidelines Governing the Protection of Privacy and Transborder Flows of Personal Data, and the Recommendation of the Council on Digital Security Risk Management for Economic and Social Prosperity, among others.

Beyond the OECD's internal framework, the Recommendation has also served as a foundational document for other significant international initiatives. Notably, the G20 Leaders welcomed the G20 AI Principles in June 2019, which were directly drawn from the OECD Recommendation. This demonstrates the Recommendation's influence in shaping global consensus on AI governance. Furthermore, the OECD AI Principles have been instrumental in informing and influencing major national and regional legislative efforts, including the European Union's AI Act. Many countries and jurisdictions have adopted the OECD's definitions of an AI system and its lifecycle in their own legislative and regulatory frameworks, fostering a degree of global interoperability and alignment in AI policy. This interconnectedness highlights the Recommendation's role as a cornerstone in the evolving architecture of international AI governance.

International Alignment

The OECD Recommendation on Artificial Intelligence stands as a pivotal instrument in fostering international alignment on AI governance, being the first intergovernmental standard on AI globally. Its adoption by the OECD Council in 2019 marked a significant moment where 42 countries initially came together to establish a common framework for responsible AI development and deployment. This number has since grown to over 46 countries, including all G7 and G20 nations, which have formally adhered to or endorsed the principles, underscoring its widespread international acceptance and the shared commitment to human-centric, trustworthy AI. The Recommendation's success in achieving such broad consensus is attributed to its balanced approach, promoting AI innovation while ensuring respect for human rights and democratic values.

The OECD actively promotes international co-operation for trustworthy AI, encouraging governments and stakeholders to work together in various global and regional fora to advance these principles. This includes fostering the sharing of AI knowledge and promoting the development of multi-stakeholder, consensus-driven global technical standards for interoperable and trustworthy AI. A key example of this international collaboration is the Global Partnership on Artificial Intelligence (GPAI), an integrated partnership that brings together OECD members and GPAI countries to advance an ambitious agenda for implementing human-centric, safe, secure, and trustworthy AI, directly embodying the principles of the OECD Recommendation. This concerted effort aims to create a stable policy environment at the international level, fostering trust in and adoption of AI across societies and economies, and ensuring that AI's transformative potential is harnessed for global benefit while addressing its challenges collectively.

Implementation Timeline

Adoption and Endorsement

Sources and References