Amazon Web Services

Success for the first Global AI Dialogue should be measured by its ability to establish a credible, multistakeholder forum where all stakeholders - including industry - participate on an equal footing in the Dialogue's discussions. In terms of outcomes, the Dialogue should surface evidence about AI adoption challenges, implementation realities, and best practices from diverse contexts, creating a shared knowledge base that informs effective policymaking. This can include identifying successful domestic policy levers that increase AI Adoption, and where international cooperation is needed. A successful Dialogue would focus on the importance of risk-based regulation rooted in international AI standards, avoiding prescriptive frameworks that could stifle innovation or prove inflexible as technologies evolve. The Dialogue's Chair's summary could include level-setting the state of AI adoption across countries using tools such as the World Bank AI Adoption Index being developed, along with best practice compilations, and evidence-based recommendations. Critically, the Dialogue must complement - not duplicate - existing international efforts.

• Safe, secure and trustworthy AI
• AI capacity-building
• Interoperability of governance approaches
• Transparency, accountability, and human oversight

These four thematic areas directly affect responsible AI adoption across diverse contexts. They address both the foundational requirements for trustworthy AI systems and the practical challenges of deploying AI in fragmented regulatory environments, while building necessary capabilities worldwide. The following provides some additional explanation. - Safe, secure and trustworthy AI is foundational to adoption and public trust. Without confidence in AI systems' safety, security, and reliability, organizations and individuals will hesitate to adopt AI technologies, limiting the realization of potential benefits. - AI capacity-building is essential to ensure that AI's benefits reach all countries and communities. - Interoperability of governance approaches stands out as a key priority because regulatory fragmentation creates concrete barriers to AI development and deployment, particularly for small and medium-sized enterprises (SMEs) and innovators in developing countries. International standards and mutual recognition frameworks can build bridges across regulatory regimes, enabling compliance while supporting innovation. - The transparency, accountability and robust human oversight of artificial intelligence systems in a manner that complies with international law are core to AWS's responsible AI strategy and essential to building and maintaining the trust necessary for broad AI adoption. AWS recognizes without clear mechanisms for accountability and human control embedded across the full AI lifecycle organizations cannot confidently deploy or rely on AI systems for meaningful work.

AI adoption emerges as the critical cross-cutting issue not explicitly mentioned in the seven thematic areas. How, where, and by whom AI is adopted ultimately determines whether the technology's development benefits are realized globally. Governance frameworks that inadvertently create barriers to beneficial AI deployment-through excessive compliance burdens, unclear requirements, or conflicting obligations-risk slowing progress in AI adoption. For example, according to a 2025 World Bank Digital Progress and Trends Report, low AI adoption amongst low-income countries (LICS) "stems not from limited usefulness but from systemic barriers that demand policy intervention. Coordination challenges prevent small firms and farmers from achieving the scale needed for collective AI investment, and information asymmetries, particularly in LICs, limit awareness of AIs applicability or access to localized models." These challenges to AI adoption underscore the importance of governance being calibrated to a country's AI readiness. Relatedly, a 2026 OECD Survey "Empowering SMEs in the age of AI", found that 13% of SMEs identified regulatory complexity as an obstacle to AI adoption. Given this context, the Dialogue should examine factors that enable or impede responsible AI adoption, including: ● Infrastructure access and affordability: Ensuring that organizations worldwide can access computing resources, data storage, and connectivity necessary for AI development and deployment. ● Regulatory clarity and predictability: Providing clear guidance that enables organizations to invest confidently in AI systems while meeting governance requirements. ● Interoperability and portability: Avoiding lock-in and enable organizations to adopt AI solutions that work across platforms and jurisdictions. ● Skills and capacity: Building the human capital necessary to develop, deploy, maintain, and govern AI systems effectively. Another emerging concern is ensuring that governance frameworks remain flexible and adaptable as AI technologies evolve. Prescriptive regulations that specify particular technical approaches risk becoming obsolete quickly or inadvertently favoring certain technologies over potentially superior alternatives. Outcome-focused, risk-based governance that allows for technological evolution offers a more sustainable approach.

Regulatory fragmentation: Conflicting requirements across jurisdictions force organizations to navigate complex, often contradictory compliance landscapes. For example, differing definitions of "high-risk" AI systems, varying transparency requirements, and incompatible conformity assessment procedures create compliance costs and delays. These burdens fall disproportionately on SMEs and developing country innovators, both of whom often lack the legal and technical resources to manage multi-jurisdictional compliance. Opportunities for international cooperation: International standards offer the most promising path to building bridges across regulatory regimes. Technical standards developed through inclusive, consensus-based processes (such as those led by ISO/IEC and NIST) can provide common frameworks for AI safety, security, transparency, and accountability that enable mutual recognition and regulatory interoperability.

The AI Dialogue should serve as a platform for evidence-sharing, coordination, and mutual learning while respecting diverse governance approaches. The key focus for the AI Dialogue should be: 1. Understanding risk-based AI regulation: The AI Dialogue should share experience on risk-based approaches to regulating AI that is also rooted in international standards. 2. Reduce fragmentation: By mapping existing national, regional, and sectoral AI governance initiatives, the Dialogue can identify areas of unnecessary divergence that create barriers to cross-border AI development and deployment. 3. Connect existing initiatives: Rather than creating new governance structures, the Dialogue should connect disparate efforts including OECD AI Principles, G7 Hiroshima AI Process, regional frameworks (EU AI Act, ASEAN guidelines, African Union AI strategy), technical standards bodies (ISO/IEC, IEEE), and sectoral initiatives. 4. Generate evidence: The Dialogue should serve as a repository and dissemination platform for evidence about what governance approaches work in practice. This includes case studies of successful AI deployments, evaluations of different regulatory approaches, and documentation of implementation challenges and solutions.

The AI Dialogue should build upon a rich ecosystem of existing AI governance initiatives, frameworks, and best practices rather than starting from scratch. This includes work on the OECD, the G7 Hiroshima AI Process, and the AI Safety Summits, Technical standards and frameworks such as are being developed in ISO/IEC JTC 1/SC 42. The AI Dialogue should connect with these various efforts, identify gaps in coverage or coordination, and facilitate knowledge transfer across initiatives. It could map areas of convergence and divergence across frameworks to identify opportunities for interoperability and mutual recognition.

Effective multistakeholder participation requires recognizing the distinct contributions different stakeholders bring and designing formats that enable substantive engagement. AWS and other companies can bring practical insight into real-world AI development, deployment challenges, implementation costs, and technical constraints. Effective formats for the AI Dialogue would include the following: - Interactive workshops over formal presentations: Minimize scripted panel discussions and formal presentations in favor of interactive formats that enable problem-solving and dialogue. Smaller breakout sessions organized by theme or use case allow for deeper engagement. - Solution-oriented discussions: Focus sessions on specific governance challenges with diverse stakeholder teams working collaboratively to develop approaches. - Case study deep-dives: Structured examination of real-world AI deployments can generate insights about what governance approaches work in practice. - Technical demonstrations: Hands-on demonstrations showing how AI systems work, their limitations, and security considerations can demystify technology for policymakers and enable more informed governance discussions. - Hybrid participation: Offer both virtual and in-person participation options to maximize inclusivity, particularly for stakeholders from developing countries or smaller organizations. - Pre-circulated materials: Distribute discussion papers, case studies, and technical background materials in advance of sessions to enable substantive dialogue. - Structured submission opportunities: Create formal channels for stakeholders to submit evidence, case studies, and technical papers that inform Dialogue discussions, ensuring that insights from those unable to attend in person are incorporated.

Small and medium-sized enterprises (SMEs) remain critically underrepresented in AI governance discussions despite being disproportionately affected by regulatory fragmentation. SMEs often lack the legal, technical, and financial resources to navigate complex, multi-jurisdictional compliance requirements that larger firms can manage. Yet SMEs drive significant innovation, create jobs, and often develop specialized AI applications for underserved markets or use cases. Resource and time constraints of most SMEs make it particularly important to encourage participation of SME-focused national and regional chambers of commerce. Broader industry participation is also needed. The AI ecosystem includes cloud infrastructure providers, semiconductor manufacturers, AI labs, software developers, systems integrators, sectoral AI users, and many others. Each brings distinct perspectives. Developing country innovators and AI practitioners need stronger representation. AI governance frameworks developed primarily by and for advanced economies may not address challenges specific to developing country contexts, such as limited infrastructure, different risk profiles, or distinct social and economic priorities. Sectoral AI users—healthcare providers, educators, farmers, manufacturers, financial services firms—should participate alongside AI developers. These organizations understand how AI is deployed in practice, what benefits it delivers, what challenges arise, and what governance requirements are workable versus burdensome. Researchers and academics from diverse disciplines (not just computer science) bring important perspectives on AI's social, economic, and cultural implications. Civil society organizations representing affected communities, workers, consumers, and marginalized groups provide essential perspectives on AI risks and impacts that may not be visible to developers or policymakers.

Real-world use case deep-dives: Select specific AI deployments (e.g., AI-assisted medical diagnosis in a developing country, AI-powered agricultural advisory services, AI fraud detection in financial services) and examine them comprehensively to ground governance discussions in concrete realities. Regulatory sandbox discussions: Facilitate sharing experiences with experimental governance approaches, including regulatory sandboxes, innovation hubs, and pilot programs. What has worked, what hasn't, and what can be scaled across markets in order to expand opportunities for SMEs looking to sell globally? Standards development workshops: Provide hands-on engagement with how technical standards are created and help policymakers understand how technical standards can support regulatory objectives, what AI standards already exist and what is being developed. Structured evidence submission: Create processes for stakeholders to submit case studies, technical papers, and implementation evidence that inform Dialogue discussions.

Risk-based regulatory approaches: Governance frameworks that calibrate requirements based on AI systems' risk levels rather than applying uniform rules to all AI applications enable innovation in lower-risk contexts while maintaining appropriate oversight of high-risk applications. Computing infrastructure access: Policies that avoid unnecessary localization requirements ensure that organizations can access globally distributed, efficient, and resilient computing resources. Protections for proprietary or confidential information: Policies that avoid forced disclosure of proprietary or confidential information support both cybersecurity and continued investment in AI innovation. Regulatory sandboxes: Controlled environments where organizations can test innovative AI applications under regulatory oversight with temporary exemptions from certain requirements can generate valuable evidence about AI systems' performance and risks. The AI Adoption Initiative (www.adopt-ai.org/) is one example of a platform that engages with government, industry and civil society in developing countries to identify steps that can be taken to support AI adoption.