School of Humanities, University of Chinese Academy of Sciences, Beijing, China

From a cross-cultural AI governance perspective, a successful Global Dialogue on AI Governance should move beyond reiterating high-level principles and instead establish processes that are operational, inclusive, and adaptable across diverse contexts. First, success would involve developing a shared procedural architecture for global governance, rather than attempting to universalize substantive norms. Given the diversity of legal systems, institutional capacities, and cultural values, governance must be structured as an ongoing process of coordination and dialogue rather than a fixed framework. Second, the Dialogue should explicitly recognize epistemic diversity as central to AI governance. Many existing frameworks implicitly reflect particular philosophical assumptions about rationality, agency, and ethics. A meaningful outcome would be the integration of diverse knowledge systems into governance processes in ways that are structurally consequential, not merely symbolic. Third, success would include advancing practical interoperability mechanisms across governance regimes. This requires coordination between international organizations, national regulators, and private actors to reduce fragmentation while preserving contextual flexibility. Finally, the Dialogue should produce a forward-looking governance roadmap that links AI to broader global priorities, including sustainable development, equity, and human dignity. This roadmap should emphasize governance as a process of co-creation across stakeholders and regions. In this sense, success lies not only in consensus, but in establishing the foundations for legitimate, adaptive, and culturally grounded global AI governance.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

From my perspective working at the intersection of law, AI ethics, and cross-cultural governance, these priorities are essential because they address both the technical and normative dimensions of AI systems. "Safe, secure and trustworthy AI" is foundational, but trust cannot be reduced to system performance alone. It must also be understood as a function of social legitimacy, which depends on how AI systems align with the values and expectations of different communities. The inclusion of "social, economic, ethical, cultural, linguistic and technical implications" is critical because AI systems operate within complex societal contexts. Governance that does not account for these dimensions risks remaining abstract and disconnected from real-world deployment. "Interoperability of governance approaches" is particularly important in a fragmented global landscape. However, interoperability should not be equated with uniformity. It should enable coordination across diverse governance models while preserving contextual adaptability. Finally, "transparency, accountability, and human oversight" are necessary to ensure meaningful control over AI systems. These principles must be implemented in ways that are accessible across jurisdictions, avoiding models that are overly technical or resource-intensive. Together, these priorities support a governance approach that is technically robust, socially grounded, and globally adaptable.

An important cross-cutting issue is the challenge of epistemic misalignment in AI systems and governance frameworks. Current approaches to AI governance often rely on implicit assumptions about knowledge, rationality, and decision-making that are not universally shared. When AI systems are deployed across different cultural and institutional contexts, this can create misalignment that cannot be addressed through technical safeguards or regulatory harmonization alone. Related to this is the need to move from representational inclusion to procedural inclusion. While diversity is frequently acknowledged, there is limited attention to how different cultural perspectives can meaningfully shape governance processes and outcomes. Another emerging concern is epistemic concentration, where a small number of actors-whether states, corporations, or research institutions-define dominant AI paradigms. This risks marginalizing alternative knowledge systems and limiting the scope of global governance. Additionally, asymmetries in governance capacity remain a critical issue. Many regions lack the institutional, technical, and regulatory resources to engage effectively in AI governance, which may reinforce existing inequalities. Addressing these challenges requires rethinking AI governance as a process of pluralistic knowledge coordination, where legitimacy is derived not only from outcomes, but from inclusive and context-sensitive processes of engagement.

From my perspective working across multiple jurisdictions and engaging with cross-cultural AI governance, the impact of current governance gaps is most visible in the disconnect between global frameworks and local realities. One of the most significant challenges is **regulatory and epistemic misalignment**. Many global AI governance models are developed in technologically advanced contexts and are later adopted or referenced elsewhere without sufficient adaptation. This creates gaps where regulatory frameworks exist in form but lack practical relevance or enforceability within local institutional and cultural settings. A second challenge is **fragmentation across governance approaches**. Diverging standards, compliance requirements, and policy priorities make it difficult for states and organizations—particularly in emerging economies—to navigate the global AI landscape. This fragmentation increases compliance burdens while limiting meaningful participation in global standard-setting processes. There is also a persistent issue of **capacity asymmetry**, where limited technical expertise, institutional readiness, and regulatory infrastructure constrain effective engagement with AI governance. This not only affects implementation but also reduces the ability of these regions to shape global norms. At the same time, these gaps create important opportunities. They allow for the development of **context-sensitive governance models** that are better aligned with local values, institutional structures, and societal needs. There is also an opportunity to move beyond passive adoption toward **active contribution**, where diverse regions help shape more inclusive and pluralistic global governance frameworks. In this context, the current moment presents a critical opportunity to shift from fragmented and externally driven approaches toward **locally grounded, globally connected AI governance systems**.

From a cross-cultural AI governance perspective, the AI Dialogue can play a pivotal role by shifting international cooperation from principle-based alignment to process-based coordination. At present, global cooperation on AI is often limited to high-level consensus around shared values, while practical coordination remains fragmented. The Dialogue can address this by serving as a neutral, multi-stakeholder platform where governments, technical communities, civil society, and industry engage in sustained and structured interaction. One key role would be to facilitate interoperability across governance approaches without requiring uniformity. Rather than harmonizing all regulatory systems, the Dialogue can support mechanisms that allow diverse models to communicate, align where necessary, and coexist where appropriate. The Dialogue can also advance cooperation by embedding procedural inclusivity. This means ensuring that participation from different regions is not merely symbolic, but that diverse perspectives actively shape governance processes and outcomes. In doing so, it can help address deeper issues of legitimacy and trust in global AI governance. Additionally, the AI Dialogue can function as a coordination hub, linking existing initiatives, reducing duplication, and identifying gaps where collaboration is most needed. It can also support the translation of global discussions into actionable guidance for national and regional implementation. Ultimately, the Dialogue's value lies in enabling a form of international cooperation that is adaptive, inclusive, and grounded in real-world governance challenges, rather than limited to abstract agreement.

The AI Dialogue should build upon existing global and multi-stakeholder initiatives, including efforts led by organizations such as the International Telecommunication Union, UNESCO, and the OECD, as well as collaborative platforms like the Global Partnership on Artificial Intelligence. These initiatives have contributed significantly to norm-setting, capacity-building, and the development of ethical and policy frameworks. Additionally, projects such as the Linking AI Principles initiative demonstrate the potential for identifying convergence across global AI ethics frameworks, while regional regulatory developments and national strategies provide important insights into implementation. However, these efforts often operate in parallel, with limited coordination across institutional, regional, and sectoral boundaries. The AI Dialogue can add value by acting as a connecting layer between these initiatives, enabling more structured exchange, alignment, and mutual learning. Its added value would lie in three areas. First, it can enhance coherence across fragmented governance landscapes by facilitating dialogue between different frameworks and actors. Second, it can strengthen inclusivity by ensuring that underrepresented regions and perspectives are meaningfully integrated into global discussions. Third, it can promote practical translation, helping bridge the gap between high-level principles and context-specific implementation. In this way, the AI Dialogue would not replace existing initiatives, but rather amplify their impact by providing a platform for integration, coordination, and collective evolution of global AI governance.

Stakeholders should contribute based on their comparative strengths: governments (regulation), industry (technical and operational insight), academia (analysis), and civil society (rights and societal impact). The Dialogue should ensure these inputs interact, not remain siloed. Structurally, a three-layer model would work best: Plenary sessions to set priorities Thematic working groups for focused discussion Regional/sectoral tracks to capture context These layers should be connected, with outputs feeding into each other. The Dialogue should also be process-driven, not a one-off event. This means iterative consultations, draft feedback cycles, and continuous engagement. Finally, participation must address capacity gaps, including support for underrepresented stakeholders and flexible participation formats. The goal is a practical, ongoing platform for coordinated AI governance.

Three groups are consistently underrepresented: Stakeholders from developing regions — often affected by AI but not shaping governance Non-technical disciplines and knowledge systems — including philosophy, social sciences, and local knowledge Implementation-level practitioners — such as public sector actors, educators, and small innovators Inclusion should move beyond representation to real influence. This can be done by: Linking regional consultations to global processes Providing capacity-building support Designing formats where diverse inputs shape outcomes, not just discussions Inclusion is not only a fairness issue — it is essential for effective and legitimate governance.

The Dialogue should prioritize interactive formats over traditional panels: Scenario-based discussions to examine real-world AI use cases Co-design labs where stakeholders jointly develop governance solutions Iterative consultation cycles to refine outputs over time Asynchronous digital participation to enable broader global input Cross-regional exchanges to share practical experiences These formats shift the Dialogue from discussion to collaboration and output-driven engagement.

Several existing approaches offer useful lessons for effective AI governance, particularly where they move from principles to implementation. First, risk-based regulatory models, such as those adopted in different jurisdictions, provide a structured way to align governance requirements with levels of potential harm. These approaches are valuable because they are scalable and adaptable across sectors. Second, multi-stakeholder frameworks developed by international organizations have helped establish shared ethical baselines and facilitate coordination. Their strength lies in inclusivity and norm-setting, though their impact depends on effective translation into practice. Third, initiatives focused on mapping and comparing AI principles globally demonstrate that convergence is possible across diverse contexts. These efforts highlight common ground while also revealing areas where local adaptation is necessary. Fourth, regulatory sandboxes and pilot programs offer a practical mechanism for testing AI systems under controlled conditions. They allow policymakers and developers to identify risks, refine standards, and build institutional capacity before large-scale deployment. Fifth, emerging practices in AI auditing and impact assessment are important for operationalizing accountability. When designed effectively, they provide structured methods to evaluate system performance, fairness, and compliance. Across these examples, a common lesson is that effective governance depends not only on well-defined principles, but on mechanisms that enable continuous adaptation, contextual flexibility, and stakeholder coordination. Moving forward, the most promising approaches are those that combine global coordination with local grounding, ensuring that governance frameworks remain both relevant and implementable.