137AI > Governance > International Coordination on AI

International Coordination on AI Governance

International coordination on AI governance addresses the cross-border dimension of a technology whose deployment, supply chain, and consequences operate across national boundaries. The category covers the institutional architecture of treaties, multilateral instruments, bilateral arrangements, and industry coordination that operates between national regulatory frameworks. The work is consequential because AI deployment operates globally even where regulation operates nationally, and the gap between deployment scale and regulatory coordination is one of the substantive governance challenges of the agentic AI era.

The page addresses the systematic treatment of coordination mechanisms. The national and regional frameworks themselves are covered in Regulatory Frameworks. The technical standards bodies that operate at international scale are covered in Standards Bodies. The specific AI-CIP intersection internationally appears in Critical Infrastructure Policy Intersection. This page is the institutional architecture and dynamics of international coordination itself.

The Architecture of International AI Coordination

International coordination operates through several distinct mechanisms with different binding effect and operational consequence.

Mechanism	Binding Effect	Operational Consequence
Treaties and binding international agreements	Legally binding on signatory states; require domestic implementation	The strongest form of international coordination; produces obligations that signatories must implement
Multilateral institutional instruments	Vary by instrument; principles documents typically not binding, decisions of treaty bodies may be	Shapes national policy through political commitment without typically binding directly
Bilateral agreements	Vary by agreement; MoUs typically not binding, executive agreements may be binding	Direct coordination between two governments on specific topics; substantial scope variance
Soft law principles	Not legally binding	Shapes practice through influence on national action, industry adoption, and reputational consequence
Industry-led coordination	Not legally binding except where contracts apply	Produces industry practice, voluntary commitments, and information sharing without government authority
Sector-specific international frameworks	Vary; some treaty-based and binding (ICAO, IMO), others soft law	Substantial in specific sectors where the framework is binding; less consequential where soft law

The architecture produces a layered landscape where different mechanisms operate simultaneously. Operators navigate the layers with attention to which carry actual obligations versus political commitments.

The Council of Europe Framework Convention on AI

The Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, opened for signature in September 2024, is the first legally binding international treaty specifically on AI. The instrument represents the most substantial international coordination achievement on AI to date.

The Convention covers AI system design, development, deployment, and use throughout the AI system lifecycle. The substantive obligations address human rights protection, transparency, accountability, oversight, equality and non-discrimination, privacy, safety and robustness, and the broader integration of AI with democratic and rule-of-law values. The framework operates at principle level with substantial flexibility for national implementation.

The signatory list includes the European Union, the United States, the United Kingdom, Canada, Israel, Japan, Norway, Iceland, Andorra, Georgia, Moldova, Montenegro, San Marino, and additional states. The combination of signatories includes most major AI-developing economies outside China and provides substantial international legitimacy to the framework.

The relationship with the EU AI Act and other national frameworks is complementary rather than replacing. The Convention operates at framework level; national implementation operates at the detail level. Signatories implement the framework through their existing or developing national approaches with appropriate adjustments.

The Convention's binding nature distinguishes it from earlier multilateral AI principles documents. Where the OECD AI Principles and similar documents shape practice through influence, the Convention creates legal obligations that signatories must implement. The practical operational consequence will become clearer as implementation proceeds and as the treaty body operates.

G7 and the Hiroshima AI Process

The G7 has been a primary venue for international coordination on AI among major economies. The 2023 G7 Hiroshima AI Process produced substantial commitments that continue to shape international AI policy.

The Hiroshima Process Code of Conduct for AI developers represents a voluntary commitment by participating organizations to address AI safety, security, and trustworthiness. The code addresses identification and mitigation of risks throughout the AI lifecycle, investment in security, public reporting on capabilities and limitations, prioritization of AI safety research, and additional substantive commitments.

The Hiroshima Process Guiding Principles articulate the broader policy framework that G7 governments commit to implementing. The principles influence national action while not binding directly.

Subsequent G7 work has built on the Hiroshima foundation through the Italian G7 Presidency, German G7 Presidency, and other rotations. The G7 has become an established venue for AI coordination among the participating economies.

G20 engagement on AI has been less developed than G7 but increasing. The broader G20 membership including major emerging economies adds important perspectives that G7 alone does not include. The G20 AI work has emphasized inclusion and development considerations alongside the safety and trust emphases of G7 work.

OECD AI Principles and Policy Observatory

The OECD AI Principles, adopted in 2019 and updated in 2024, provide the foundational soft law instrument for AI governance. The Principles have been influential despite their non-binding nature because OECD work substantially shapes national policy across member economies and beyond.

The Principles cover inclusive growth and well-being, human-centered values and fairness, transparency and explainability, robustness and security, and accountability. The framework has been broadly adopted including by non-OECD economies and provides the foundation that subsequent AI coordination work has built on.

The OECD AI Policy Observatory maintains comparative data on national AI policy across approximately 70 countries. The Observatory provides systematic information about national AI strategies, regulatory frameworks, and policy initiatives that supports coordination and benchmarking.

The OECD AI Wonk and broader OECD AI work continue to develop the framework through expert engagement, comparative analysis, and ongoing principle refinement. The OECD operates as the closest existing institution to a comprehensive international AI policy coordination body, with substantial influence even where it lacks regulatory authority.

UN Coordination

UN-level coordination on AI has been developing through several channels with substantial activity but limited binding outcomes.

The UN AI Advisory Body, established in 2023, produced substantial reports including the 2024 "Governing AI for Humanity" final report. The Body's recommendations address international AI governance gaps and propose coordination mechanisms including a global AI office. The recommendations are advisory and their implementation depends on subsequent action by member states.

The Global Digital Compact, adopted at the September 2024 Summit of the Future, includes substantial AI provisions. The Compact addresses AI safety, ethics, capacity building, and international cooperation through soft law commitments by participating states. The implementation pace will determine the Compact's operational consequence.

UNESCO has produced the Recommendation on the Ethics of Artificial Intelligence (2021) covering broad ethical principles for AI. The Recommendation has been adopted by UNESCO member states and provides a multilateral framework with broader geographic reach than OECD-centered work.

UN specialized agencies including the ITU (telecommunications), WIPO (intellectual property), and others engage AI within their respective mandates. The agency-specific work addresses technical and operational dimensions that broader UN coordination does not specifically cover.

UN Security Council engagement on AI has been intermittent. The Council has held briefings on AI implications for international peace and security; substantive action has been limited.

The aggregate UN coordination produces substantial policy framework material with limited binding effect. The instruments influence national action through political commitment and reputational consideration rather than legal obligation.

The AI Safety Summit Series

The AI Safety Summit series initiated by the United Kingdom at Bletchley Park in November 2023 has become an established venue for international coordination on frontier AI safety.

The Bletchley Park Summit produced the Bletchley Declaration signed by approximately 28 countries including the EU, US, UK, China, India, Brazil, and additional major economies. The Declaration acknowledged frontier AI risks and committed to international coordination on safety research. The combination of signatories including both Western democracies and major non-Western economies including China provided unusual breadth of international engagement.

The Seoul AI Safety Summit in May 2024 continued the series with substantial focus on frontier model safety, AI Safety Institute coordination, and the broader infrastructure for AI safety work. The Summit produced the Seoul Declaration and additional commitments by participating states.

The Paris AI Action Summit in February 2025 emphasized AI for global challenges alongside continued safety work. The Summit broadened the agenda beyond frontier safety to include AI for development, sustainability, and broader social benefit.

The Summit series has produced an established network of AI Safety Institutes across participating countries. The UK AI Safety Institute, US AI Safety Institute, Japan AI Safety Institute, and equivalent bodies in additional countries coordinate through specific arrangements including the MoU between US and UK AISIs.

The Summit series operates as soft law coordination but has produced substantive institutional development. The AI Safety Institute network provides specific cross-border coordination on frontier AI evaluation and safety research that did not exist before the series began.

Bilateral and Plurilateral Mechanisms

Bilateral and plurilateral arrangements provide direct coordination between specific governments on specific topics. The mechanisms have been more operationally productive than multilateral frameworks for specific coordination needs.

The US-EU Trade and Technology Council (TTC) has been a substantial forum for transatlantic AI coordination. The TTC working groups address AI, data flows, technology standards, and other technology policy topics. The TTC has produced specific deliverables including the Joint Roadmap for Trustworthy AI and Risk Management.

The US-UK Memorandum of Understanding on AI Safety Institute cooperation establishes formal coordination between the two AI Safety Institutes. The MoU enables joint evaluation work, shared research, and coordinated approach to frontier model safety.

The US-Japan AI coordination through specific bilateral arrangements addresses both safety and economic dimensions. The US-Korea AI cooperation operates through similar arrangements with specific focus on AI safety and semiconductor coordination.

EU bilateral arrangements with the UK, Japan, Korea, and other partners operate through trade and cooperation frameworks with AI-specific elements. The arrangements support coordinated approach across the EU and its key partners.

The Quad (US, Japan, India, Australia) AI cooperation addresses both safety and economic dimensions among the four governments. The arrangement is plurilateral rather than fully multilateral but covers substantial AI development capacity.

The bilateral and plurilateral path has produced more operational coordination than multilateral channels because the smaller scale supports more substantive engagement. The trade-off is reduced legitimacy and inclusion compared to multilateral frameworks.

Sector-Specific International Coordination

Sector-specific international frameworks have substantial history and continue to develop AI-specific dimensions.

The United Nations Economic Commission for Europe (UNECE) WP.29 framework for vehicle regulations includes UN-R 155 for connected vehicle cybersecurity and UN-R 156 for software update management. The frameworks apply in countries following UN vehicle regulations including the EU, Japan, Korea, and many additional jurisdictions. The frameworks reach AI components in connected vehicles directly.

The International Civil Aviation Organization (ICAO) addresses aviation including AI in aviation systems. The framework operates through treaty (the Chicago Convention) with substantial binding effect across signatory states. AI-specific aviation coordination has been developing through ICAO working groups and the broader aviation safety framework.

The International Maritime Organization (IMO) addresses maritime safety and security including emerging AI dimensions. The IMO operates through treaty with substantial binding effect across signatory states.

The International Medical Device Regulators Forum (IMDRF) coordinates international medical device regulation including AI/ML in medical devices. The forum is voluntary but produces substantial harmonization through participation by major medical device regulators.

The Financial Stability Board addresses financial stability dimensions of AI through specific work on AI in financial services. The FSB recommendations influence financial regulators globally.

The Bank for International Settlements engages central bank coordination on AI in financial services through the BIS Innovation Hub and related work. The coordination shapes central bank approaches to AI without binding directly.

The aggregate sector-specific landscape provides substantial international coordination in specific domains with binding effect that horizontal AI coordination has not achieved. The sector path has historically been more productive than horizontal coordination for substantive international rules.

Trade Dimensions

International trade frameworks intersect with AI governance in several specific ways.

The World Trade Organization addresses AI primarily through general trade rules rather than AI-specific provisions. The General Agreement on Trade in Services, the Agreement on Trade-Related Aspects of Intellectual Property Rights, and other WTO instruments reach AI-related trade with substantial implications for digital services regulation.

Digital trade agreements including the US-Mexico-Canada Agreement (USMCA), the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP), and the Digital Economy Partnership Agreement (DEPA) include AI-relevant provisions on data flows, source code protection, and digital products. The agreements bind signatories on these specific dimensions.

Export controls on AI technology operate through national authority with substantial international coordination. The US Export Administration Regulations control specific AI-relevant technologies including advanced semiconductors and certain AI capabilities. The EU dual-use regulation provides similar framework for EU member states. International coordination through the Wassenaar Arrangement and other multilateral export control regimes addresses some AI-related items.

Investment screening for AI-related foreign investment operates through national frameworks (CFIUS in the US, FDI screening regulations in EU member states, equivalent frameworks elsewhere) with substantial coordination among major economies. The frameworks affect cross-border AI investment and operator structure.

Sanctions and AI-relevant restrictions reach specific entities and jurisdictions. The frameworks affect AI vendor operations across affected jurisdictions and require operator compliance discipline.

The Brussels Effect

The Brussels effect describes the dynamic by which EU regulation influences global practice beyond its formal jurisdictional reach. The dynamic has been substantial for AI through the EU AI Act and related instruments.

Operators serving the EU market must comply with EU regulation. The market is substantial enough that compliance is operationally required for many AI vendors. Implementing EU-compliant practice for EU operations often makes operational sense even where operators could legally operate differently in other jurisdictions.

EU regulation often becomes the global default for vendor practice. Producing one set of products that complies with EU regulation is operationally simpler than producing different products for different markets. The default extends EU regulatory influence well beyond formal jurisdictional reach.

Other jurisdictions adopt or adapt EU frameworks. The GDPR template has been adopted with variations across many jurisdictions globally. The EU AI Act framework is influencing similar adoption patterns in other jurisdictions, though the adoption is at earlier stage given the Act's recent enactment.

The Brussels effect has limits. Substantial economies including the US and China operate AI under frameworks substantially different from the EU approach. The diversity persists despite the Brussels effect dynamic, particularly where the alternative frameworks reflect deliberate policy choices about innovation, competition, and government role.

The dynamic produces specific consequences for operators. Multi-jurisdiction operators face the choice between unified compliance to the most stringent applicable framework (often the EU) or differentiated practice across jurisdictions. The choice has substantial operational implications.

Industry-Led International Coordination

Industry coordination operates alongside government coordination with substantive influence on practice even without regulatory authority.

The Frontier Model Forum brings together major AI developers (Anthropic, Google, Microsoft, OpenAI, and additional members) for coordination on frontier AI safety. The Forum produces shared work on safety research, vulnerability disclosure, and best practice. The coordination operates as industry self-governance with limited transparency to outside observers.

The Partnership on AI is a multi-stakeholder body with broader membership including civil society, academic, and industry participants. The Partnership produces work on responsible AI development across substantive topics and provides cross-sector coordination that single-industry bodies do not.

The MLCommons benchmark coordination addresses AI evaluation including safety evaluation. The benchmarks support comparative assessment of AI systems and contribute to the broader infrastructure for AI safety work.

Industry consortia for specific applications including the C2PA for content provenance, the Coalition for Health AI, and others operate sector-specific coordination on AI practice.

Academic-industry partnerships through institutes including Stanford HAI, MIT CSAIL, Oxford Future of Humanity Institute, Cambridge Centre for the Study of Existential Risk, and others produce substantive coordination on AI research and policy. The partnerships influence both industry practice and government policy through expert engagement.

The aggregate industry-led coordination produces substantive practice and shared infrastructure that government coordination alone would not achieve. The coordination has its own legitimacy concerns including representation, transparency, and accountability that civil society and government observers raise.

The Fragmentation Problem

International coordination on AI has produced substantial activity but has not resolved fundamental fragmentation in how AI is governed across jurisdictions.

Major economies operate AI under substantially different approaches. The EU framework emphasizes comprehensive horizontal regulation with substantial precautionary discipline. The US framework is sectoral with substantial state-level variation and lighter horizontal regulation. The Chinese framework combines aggressive AI deployment with substantial state control over AI services. Other economies position along different points on the regulatory spectrum.

The fragmentation produces operational complexity for AI vendors operating internationally. The same product faces substantially different regulatory treatment across jurisdictions, requiring either differentiated products or unified compliance to the most stringent applicable framework.

The fragmentation also reflects substantive disagreement about AI governance principles. Different jurisdictions reach different conclusions about innovation versus precaution, individual rights versus collective interest, market versus government role, and democratic accountability versus efficiency. The disagreement is not merely technical and is unlikely to be resolved through coordination mechanisms alone.

Geopolitical tension between major economies particularly between the US and China shapes the AI governance landscape. Export controls, technology restrictions, and broader strategic competition affect coordination prospects. Joint coordination becomes more difficult when underlying geopolitical relationships are tense.

The coordination work continues despite the fragmentation. Bilateral arrangements, sector-specific frameworks, and industry coordination continue to develop. The progress is incremental and uneven, and the resolution of fundamental fragmentation is unlikely in the near term.

Practical Implications for Operators

For operators deploying AI internationally, the coordination landscape produces several practical implications.

Multi-jurisdiction compliance is the operational baseline. Operators with international reach face multiple regulatory frameworks simultaneously and must navigate compliance with the most stringent applicable requirements or implement differentiated practice by jurisdiction.

Treaty and convention engagement affects operator commitments. The Council of Europe Framework Convention, OECD AI Principles, and similar instruments produce expectations that operators address even where not directly binding.

Bilateral framework engagement supports specific coordination needs. Operators with substantial presence in specific bilateral relationships engage the corresponding frameworks through industry channels and direct government engagement.

Standards adoption supports compliance across multiple frameworks. International standards through ISO/IEC and similar bodies, harmonized standards under specific frameworks, and broader standards-based practice produce coherence across jurisdictional differences.

Geopolitical risk assessment addresses the strategic dimension of international AI deployment. Export controls, investment restrictions, and broader strategic considerations shape what operators can do in specific jurisdictional combinations.

Industry coordination participation supports both substantive practice and reputational positioning. Engagement in Frontier Model Forum, Partnership on AI, and equivalent industry bodies addresses both substantive coordination needs and broader stakeholder relationships.

The Reframe

International coordination on AI governance is substantial and growing but operates under significant constraints from underlying jurisdictional differences and geopolitical tension. The architecture includes treaties (most prominently the Council of Europe Framework Convention), multilateral instruments (G7 Hiroshima Process, OECD AI Principles, UN coordination including the Global Digital Compact), the AI Safety Summit series, bilateral and plurilateral mechanisms, sector-specific international frameworks (UNECE, ICAO, IMO, IMDRF, FSB), trade dimensions, and industry-led coordination. The Brussels effect produces substantial influence of EU regulation beyond its formal reach. The fragmentation problem persists because major economies reach different substantive conclusions about AI governance that coordination cannot bridge through process alone. For operators, multi-jurisdiction compliance is operational baseline with treaty engagement, bilateral framework engagement, standards adoption, geopolitical risk assessment, and industry coordination participation all part of mature practice. The coordination work continues to develop, with incremental progress through bilateral and sector-specific channels more productive than comprehensive multilateral frameworks. The pace is substantially slower than AI deployment, and the coordination gap is one of the substantive governance challenges the agentic AI era requires sustained work to address.

Related Coverage

Governance | Regulatory Frameworks | Standards Bodies | Critical Infrastructure Policy Intersection