OECD

The OECD is pleased to actively support and advance the objective of the Global Dialogue on AI Governance to build on on-going efforts to advance trustworthy AI worldwide, including by helping to promote interoperability and compatibility of AI governance approaches, while avoiding duplication. As an inclusive space to deliberate on today's most pressing AI challenges, the Dialogue would succeed by: 1. Advancing a shared understanding of foundational concepts and promote coherence and interoperability: develop a shared understanding of how the Dialogue complements and connects with existing international efforts while avoiding duplication, including by drawing on and amplifying internationally agreed references, including on: o AI systems and their lifecycle: The OECD definition of an "AI system" and the AI lifecycle are already reflected in many national and regional governance approaches. The OECD AI Systems Classification Framework allows policy makers to distinguish between AI applications, according to their potential impact on individuals, society and the planet. o AI incidents: The OECD common reporting framework for AI incidents, building on the OECD's definition of an "AI incident", promotes consistency in identifying, analysing, and learning from AI-related risks and harms, particularly as countries are exploring AI incident reporting and response frameworks. o Emerging AI developments: Recent OECD definitions analysis, including for the agentic AI landscape help governance discussions keep pace with emerging AI and policy developments. 2. Grounding its deliberations in robust evidence, built on rigorous measurement and analysis: the OECD.AI Policy Observatory provides an expansive and ever-growing knowledge-base of real-time quantitative data spanning AI research, investment, jobs, AI incidents, and emerging global AI trends. These resources help policymakers stay informed and reach a shared understanding of trends in AI on which to ultimately base their strategic choices regarding which key issues global efforts in AI policy should focus on.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

The selection reflects the OECD's potential contribution to the thematic clusters proposed by the co-Chairs' draft note on themes and structure: The advancement of safe, secure and trustworthy AI is core to the OECD AI Principles that underpin the OECD's work on artificial intelligence. They guide AI actors in their efforts to develop trustworthy AI and provide policymakers with recommendations for effective AI policies. They include provisions on transparency (OECD AI Principle 1.3), accountability (1.5) and human oversight (1.2), which the OECD implements through practical governance mechanisms, translating high-level principles into practice, including through: o The Hiroshima AI Process Reporting Framework, which encourages transparency and accountability among organisations developing or deploying AI systems; o The OECD Catalogue of Tools and Metrics, which brings together more than 1,000 helpful approaches, mechanisms and practices to help make AI safer, more secure, or more trustworthy, including open-source tools o The OECD Due Diligence Guidance for Responsible AI, which provides a framework for enterprises to set up internal management systems capable of proactively identifying and responding to risks, including those related to human rights o The forthcoming OECD AI Policy Toolkit to support economies in harnessing AI opportunities and realising the benefits of AI will provide practical guidance for countries with different capacities, tailored to specific economic and regional contexts, and to different sectors and industries. For governments to work together to manage AI on an international level, they need to use common terms and definitions to act as a foundation for cooperation. The OECD facilitates interoperability and alignment across AI governance approaches through policy frameworks and policy exchange. The OECD informs policy deliberations with insights on social and economic implications of AI and key technical developments, including on emerging AI capabilities, and the benefits and risks of open AI models.

• The theme relating to "social, economic, ethical, cultural, linguistic and technical implications of AI" would capture most potential issues. More focused discussions may help orient the dialogue toward pressing matters where international exchange could prove particularly fruitful. The Dialogue could leverage recent OECD analysis on specific aspects of AI systems to inform deliberations with insights on key technical concepts and emerging capabilities, such as:' • On the agentic AI landscape: recent OECD analysis supports a clearer conceptual understanding to inform research and policymaking and providing descriptive data on recent trends in the uptake of AI agents and agentic AI
• forthcoming work will provide insights on practical applications of agentic AI from leading companies. • On the concept of AI openness: recent work provides greater clarity on terminology, and supports policy discussions on how to balance the openness of generative AI foundation models with responsible governance
• forthcoming analysis will highlight the economic benefits of open-source frameworks in the AI stack, from optimising physical infrastructure to enabling advanced, AI-driven business solutions, economic impacts, as well as effects on innovation, local value creation and sector-specific adaptation. The Dialogue could advance the adoption of AI tools to generate benefits across a number of sectors, informed by OECD analysis, including: • On Governing with AI, where the OECD is building an up-to-date, comprehensive evidence base of AI efforts in government • On productivity gains from AI, both at firm- and at macro-level, specifically in LICs and LMICs, as well as considering the role of trade • On AI's role in finance, including in specific regions • On scaling AI in health, which includes a policy checklist to help establish strategies and action plans • On AI's impact on the future of work, including with regard to job quality, safety • On effective uses of generative AI in education, in the recent OECD Digital Education Outlook 2026

Differences in the maturity, scope, and implementation of AI governance frameworks across countries and regions create both challenges and opportunities. Fragmentation in definitions, risk classification, and reporting requirements can increase compliance costs for developers and users of AI systems, complicate cross-border deployment, and reduce the effectiveness of risk mitigation efforts. Inconsistent approaches may also slow the diffusion of beneficial AI applications, particularly for small firms and public-sector users. At the same time, the rapid development of national and regional AI strategies and initiatives (as evidenced by the more than 2,200 initiatives in the OECD.AI Policy Navigator) provides valuable opportunities for peer-learning and interoperability. Advances in areas such as AI incident reporting, transparency mechanisms, and policy evaluation are generating practical insights that can inform international cooperation. For many emerging and developing economies, governance gaps are closely linked to capacity constraints, including limited access to connectivity, data, skills, compute resources, and policy tools, and financing. Addressing these gaps presents an opportunity to support inclusive AI adoption and enhance productivity.

As a general-purpose digital technology, the development, deployment, and use of AI cut across sectors and impact countries around the world. The growing recognition of the benefits and risks of AI has led to new international initiatives on AI, involving different sets of countries and stakeholders. Many initiatives already play important and complementary roles in advancing trustworthy, human centred AI. The global opportunities and challenges presented by AI require inclusive engagement and multi-stakeholder participation. The Dialogue can thus strengthen international cooperation by connecting and amplifying existing efforts, promoting coherence and raising synergies, as well as avoiding duplicative efforts. Its value lies in broadening existing policy deliberations to a wider global audience, supporting policymakers around the world in benefitting from existing definitions, data and expert insights, and enhancing the reach of existing frameworks to further promote safe, secure and trustworthy AI and good practices across jurisdictions.

The Dialogue could highlight key fora for exchange among policymakers and AI specialists, and encourage further exchange, synergies, and participation from around the world. The Dialogue can thus help translate technical and policy learnings into broader international policy understanding, encourage participation from countries around the world, and promote interoperability across governance approaches. The Global Partnership on Artificial Intelligence (GPAI) is an international initiative that promotes trustworthy AI, grounded in the OECD AI Principles. Through its broad government membership (46 member countries across six continents, and the European Union) and expansive AI expert community, GPAI addresses global AI challenges by combining technical expertise with robust governance frameworks and building bridges between research and AI policy development. GPAI collaborates with the United Nations and other international organisations and contributed to the international AI Summit series, most recently through thematic events on transparency, open-source tools and AI in agriculture, as well as a Ministerial meeting of the GPAI at the India AI Impact Summit. A key forthcoming output of GPAI's work will be an AI Policy Toolkit, co-created with countries across Africa, Asia and Latin America, to help governments design and deploy trustworthy AI suited to local contexts. The Hiroshima AI Process (HAIP) Reporting Framework: As part of the G7 Hiroshima AI Process, the OECD, informed by leading AI developers, supported the G7 in developing the HAIP reporting framework to encourage transparency and accountability among organizations developing advanced AI systems and facilitate the application of the Hiroshima AI Process International Code of Conduct. The first round of submissions (25 reports by 9 organisations in 2025) shows the reporting framework has already helped improve coordination across teams, enabled benchmarking against international standards, and clarified internal communication on AI governance. Testimonials also indicate that participating companies have derived benefit vis-à-vis clients and investors alike. Based on stakeholder feedback during the first year of implementation, a streamlined version is being finalised.

The OECD's experience in developing the OECD AI Principles in 2019 and now with the broad community of more than 500 experts supporting the Global Partnership on Artificial Intelligence, points to the importance of involving stakeholders in AI policy design. Institutional stakeholders from industry, trade unions, the technical community and civil society take part in the OECD's intergovernmental meetings on AI, representing the voice of their members, which helps the OECD draw on the latest developments in the AI industry, to include the latest technical considerations and research from academia, and to benefit from the voices to civil society. Representatives from other international organisations are invited to contribute to OECD deliberations on a regular basis. The expansive, multidisciplinary and multi-stakeholder GPAI expert community with more than 500 AI specialists is organised in expert groups which contribute to specific analytical outputs on key AI issues – such as AI incidents, agentic AI, AI compute or AI investment. It is essential that the Dialogue involves a broad group of stakeholders from different backgrounds as well as other international organisations outside the UN system. The OECD looks forward to being involved and to contributing its tools and insights.

Inclusion is helpful but meaningful participation should be the end-goal. Meaningful participation requires the capacity and knowledge to engage and is a pre-requisite to inclusion yielding real outcomes. Many regions and countries are underrepresented in global discussions on AI governance, but would be best placed for developing context relevant AI systems - since these same regions and countries are likely to be strongly impacted by AI developments. The UN and others, including the OECD, must ensure that these impacts are positive – for example by improving healthcare, education, infrastructures, and overall well-being. Doing so will take concerted and pro-active engagement that should include development efforts (including by development banks and bilateral aid agencies) and statistical capacity building measures for communities, as well as local businesses, academic institutions and civil society groups.

Innovative engagement formats have been an essential component of the development of the forthcoming OECD AI Policy Toolkit, which aims to support economies in harnessing the benefits of AI and help them implement the OECD AI Principles in a timely and effective manner through practical, non-prescriptive guidance that matches national contexts, capacities, and priorities. Co-creation workshops involving stakeholders from countries across Africa, Asia, Europe, and Latin America and the Caribbean respectively helped identify challenges in designing and implementing AI policies, relevant policy examples and resources needed to improve AI policy readiness. More broadly, to be meaningful, useful and credible, engagement formats must be tailored to their purpose. Context-specific and goal-driven engagement – such as the OECD expert groups which have a clear focus on a topic and deliverable – with experts whose views are well-respected and insightful and stakeholder representatives which bring together the perspectives of different groups, can foster broad, meaningful participation and impactful outputs. Other than the OECD formats, multistakeholder groups like the community of the Internet Corporation for Assigned Names and Numbers (ICANN) and the Internet Governance Forum (IGF), among others, have managed to combine broad participation with overall meaningful and actionable outputs.

The OECD.AI Policy Observatory is an online interactive platform dedicated to promoting safe, secure and trustworthy AI. Since 2020 the Observatory has been an essential resource for policymakers, researchers, businesses, and civil society, offering a comprehensive view of global AI initiatives, trends, and governance frameworks. OECD.AI offers a wide range of tools and data which could support the AI Dialogue by providing a shared empirical foundation for policy dialogue, enabling comparative learning, and tracking progress over time, including: • The OECD.AI Policy Navigator - a government-vetted repository of over 2,200 AI policy initiatives from more than 88 jurisdictions and organisations - enables comparative analysis and supports identification of AI governance gaps, lessons and good practices. • Live data dashboards on AI investments, research outputs, and policy developments and interactive visualisations help users compare and assess global progress in AI governance and analyse AI trends in real-time. • On the Catalogue of Tools & Metrics, AI practitioners from around the world can share and compare tools, build on each other's efforts, and create global best practices to advance trustworthy AI. The catalogue allows users to submit their experiences as use cases, where they can give guidance and insights. • The OECD AI Incidents and Hazards Monitor leverages large language models to document AI incidents and hazards, helping policymakers, AI practitioners, and all stakeholders worldwide gain valuable insights into the risks and harms of AI systems. Over time, AIM will help to show risk patterns and establish a collective understanding of AI incidents and hazards and their multifaceted nature, serving as an important tool for trustworthy AI. • The upcoming AI Policy Toolkit platform will be an interactive platform helping governments gauge their level of alignment with OECD AI Principles in order to better prioritise and target their implementation efforts, practical, non-prescriptive guidance that matches national contexts, capacities, and priorities.