Università della Svizzera italiana (USI) / International Human Rights and Justice Institute (IHRJI)

A successful first Global Dialogue on AI Governance would move beyond high-level discussion and produce clear, actionable priorities that can guide real-world implementation. It should identify practical principles around transparency, accountability, and due process that are robust enough to apply across jurisdictions and sectors, rather than remaining abstract commitments. From the perspective of ethics, compliance, and dispute resolution, success would also mean recognizing that AI-related conflicts are inevitable. The dialogue should begin addressing how such disputes can be prevented and resolved, including issues of liability, evidentiary standards, and the role of arbitration and other cross-border mechanisms. Without this, governance risks remaining disconnected from enforcement realities. Meaningful inclusion is another essential outcome. Perspectives from developing countries, as well as practitioners working in regulatory, legal, and compliance fields, should actively shape the agenda. Otherwise, governance frameworks risk being overly concentrated and impractical in broader global contexts. The dialogue should also lead to concrete next steps, such as establishing working groups or pilot initiatives aimed at harmonizing approaches to AI governance and dispute resolution across jurisdictions. Continued engagement will be critical to maintaining momentum. Finally, success would be reflected in building trust and legal certainty. As AI systems increasingly operate across borders, predictable and fair governance frameworks are essential for both innovation and accountability. Overall, the first dialogue should not attempt to resolve all challenges, but should establish a credible foundation for coordinated, enforceable, and globally relevant AI governance.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Safe, secure and trustworthy AI
• Interoperability of governance approaches

My selection focuses on areas where AI governance needs to become practical, especially across borders. Transparency, accountability, and human oversight are critical because without them it is difficult to understand how systems operate or who is responsible when something goes wrong. This creates real challenges for compliance and dispute resolution. The protection and promotion of human rights is also central. Many of the legal and regulatory issues around AI ultimately come down to rights-based concerns, which makes this a key area for both governance and enforcement. Interoperability of governance approaches is important given how widely AI systems are used across jurisdictions. When regulatory approaches diverge too much, it increases uncertainty and makes cross-border disputes harder to manage. Some level of alignment would improve predictability. Finally, safe, secure and trustworthy AI underpins all of the above. Governance efforts will only be credible if the systems themselves are reliable and do not introduce unnecessary risks. Overall, these priorities reflect the need for approaches that are workable in practice, not just well-defined in theory.

One cross-cutting issue that deserves more attention is the question of legal responsibility and dispute resolution in AI-related harms. While accountability is mentioned, there is still limited clarity on how liability should be assigned when AI systems operate across multiple actors, jurisdictions, and layers of decision-making. This creates uncertainty not only for regulators, but also for businesses and individuals seeking redress. A related emerging issue is the growing role of AI in decision-making processes that have legal or quasi-legal consequences, including areas such as finance, employment, and even dispute resolution itself. This raises questions about due process, evidentiary standards, and the extent to which automated outputs can or should be relied upon in formal proceedings. Another gap is the operationalization of governance. Many frameworks focus on principles, but there is less emphasis on how these principles are implemented, monitored, and enforced in practice. Without clear mechanisms for oversight and compliance, governance risks remaining largely aspirational. Finally, the concentration of technological power remains an underlying concern. A small number of actors continue to shape the development and deployment of advanced AI systems, which has implications for fairness, access, and global influence. These issues cut across existing themes and highlight the need to connect governance discussions more closely with legal, institutional, and enforcement realities.

Governance gaps in areas such as transparency, accountability, and interoperability are already creating tangible challenges in cross-border and sector-specific contexts. In practice, the lack of clear and consistent standards makes it difficult to assess responsibility when AI systems are involved in decision-making processes. This is particularly visible in international dispute resolution, where questions around liability, evidentiary reliability, and procedural fairness are becoming more complex. Diverging regulatory approaches across jurisdictions further increase uncertainty. For actors operating internationally, this fragmentation complicates compliance and raises the risk of conflicting obligations. It also makes dispute resolution more difficult, as there is no common baseline for evaluating AI-related conduct or harm. At the same time, these developments create opportunities. AI has the potential to improve efficiency in compliance processes, enhance risk assessment, and support more informed decision-making in legal and regulatory contexts. In dispute resolution, AI tools may assist with case analysis, document review, and procedural management, provided their use is transparent and properly governed. However, without stronger alignment on governance approaches and clearer mechanisms for accountability, these benefits remain uneven and potentially contested. The current landscape reflects a tension between rapid technological adoption and slower regulatory adaptation. Overall, the most significant challenge lies in bridging this gap between innovation and enforceable governance, while the key opportunity is to shape frameworks that make AI both usable and accountable across jurisdictions.

The AI Dialogue can play a meaningful role by acting as a bridge between high-level principles and practical coordination. Many existing initiatives focus on setting norms, but there is still a gap when it comes to aligning how these are interpreted and applied across jurisdictions. The Dialogue can help reduce this gap by fostering a shared understanding of key governance concepts, particularly in areas such as accountability, transparency, and risk management. It can also serve as a platform for identifying points of convergence between different regulatory approaches. Rather than attempting full harmonization, which is often unrealistic, the Dialogue can promote interoperability by encouraging compatible standards and mutual recognition where possible. This would be especially valuable for cross-border activities, where inconsistent rules create legal uncertainty and complicate compliance. Another important role is to bring together not only policymakers, but also practitioners from legal, technical, and compliance fields. Their involvement can help ensure that governance discussions are grounded in operational realities, including how disputes arise and are resolved in practice. The Dialogue can further support international cooperation by initiating targeted follow-up actions, such as working groups or pilot initiatives focused on specific challenges. This would help maintain momentum and move discussions toward implementation. Ultimately, its value will depend on whether it contributes to building trust and predictability across jurisdictions. If it can facilitate more coherent and workable approaches to AI governance, it will play a constructive role in advancing international cooperation.

The AI Dialogue should build on existing initiatives that have already developed principles, standards, and policy coordination mechanisms. These include efforts by the OECD on AI principles, the UNESCO Recommendation on the Ethics of AI, and the Global Partnership on AI, which brings together governments and experts to advance responsible AI. Regional regulatory frameworks, such as the EU AI Act, also provide important reference points for implementation. In addition, standard-setting bodies like ISO and IEEE play a key role in translating high-level principles into technical standards. These efforts collectively form a strong foundation, but they remain fragmented across regions, sectors, and levels of application. The added value of the AI Dialogue lies in its ability to connect these initiatives and promote greater coherence. Rather than duplicating existing work, it can act as a coordination platform that helps align interpretations, identify common priorities, and encourage interoperability across governance approaches. It can also bring together a broader range of stakeholders, including practitioners in legal, compliance, and dispute resolution fields, to ensure that governance discussions reflect operational realities. This would help bridge the gap between principles and enforceable mechanisms. Ultimately, the Dialogue can contribute by reinforcing synergies between existing efforts and supporting more consistent, predictable, and globally relevant approaches to AI governance.

Different stakeholders can contribute to the AI Dialogue by focusing on their comparative strengths. Governments can provide policy direction and ensure alignment with public interest objectives. The private sector can share technical expertise and implementation experience, particularly around system design, risk management, and operational constraints. Civil society and academia can contribute independent analysis, highlight societal impacts, and help ensure that governance approaches remain inclusive and rights-based. Practitioners in legal, compliance, and dispute resolution fields can bring practical insight into how rules are applied, enforced, and challenged in real-world contexts. To be effective, the Dialogue should be structured to move beyond general statements toward more focused and outcome-oriented discussions. A combination of plenary sessions and smaller, thematic working groups would allow for both broad engagement and more detailed exchanges. Working groups could focus on specific issues such as accountability, cross-border interoperability, or dispute resolution mechanisms. The format should also encourage interaction rather than one-way presentations. Case-based discussions, scenario analysis, and problem-solving sessions would help ground the dialogue in practical challenges. Including follow-up mechanisms, such as ongoing working groups or periodic reporting, would be important to maintain continuity and track progress. Finally, the Dialogue should remain open and inclusive, while ensuring that contributions are structured and purposeful. Clear agendas, defined outputs, and opportunities for diverse stakeholders to engage meaningfully will be key to making the process both credible and effective.

Several voices remain underrepresented in global discussions on AI governance. Perspectives from developing countries, particularly from regions in the Global South, are still limited, despite being significantly affected by the downstream impacts of AI systems. This creates a risk that governance frameworks reflect the priorities of a narrow group of technologically advanced actors. Practitioners working in implementation and enforcement contexts are also often overlooked. This includes professionals in compliance, legal practice, dispute resolution, and regulatory oversight, who deal directly with how AI-related rules are applied and challenged in practice. Their absence can result in frameworks that are difficult to operationalize. In addition, affected communities, including workers impacted by automation, individuals subject to algorithmic decision-making, and smaller businesses with limited resources, are not consistently represented. Their experiences are critical for understanding real-world consequences beyond high-level policy discussions. To address these gaps, inclusion needs to go beyond formal representation. This could involve targeted outreach, financial and logistical support to enable participation, and structured consultation processes that allow these groups to contribute meaningfully. Regional dialogues and partnerships with local institutions can also help ensure that diverse perspectives are reflected. Finally, the Dialogue should integrate practical expertise alongside policy discussions, ensuring that those involved in implementation and dispute resolution have a role in shaping governance approaches.

To foster meaningful and dynamic engagement, the AI Dialogue should move beyond traditional panel discussions and adopt more interactive, problem-focused formats. One effective approach would be case-based simulations, where participants work through realistic cross-border AI scenarios, such as disputes involving liability, bias, or regulatory conflict. This encourages stakeholders to engage with practical challenges and understand different perspectives. Multi-stakeholder breakout labs could also be valuable. Small, diverse groups could focus on specific issues, such as accountability or interoperability, and be tasked with producing short, concrete outputs. This format promotes deeper discussion and avoids repetitive high-level statements. Another option is scenario stress-testing, where proposed governance principles or frameworks are tested against complex, real-world situations. This helps identify gaps between theory and implementation and encourages more robust solutions. The Dialogue could also include structured debates, where participants are assigned contrasting positions on contentious topics. This can surface assumptions, clarify disagreements, and prevent superficial consensus. Finally, ongoing digital collaboration platforms can extend engagement beyond in-person sessions. These platforms could support iterative input, document sharing, and follow-up discussions, ensuring continuity and broader participation. Overall, the most effective formats will be those that prioritize interaction, practical problem-solving, and clear outputs, rather than passive listening. This shift is essential if the Dialogue is to generate insights that are both credible and actionable.

Several existing policies and practices provide useful foundations for effective AI governance. The OECD AI Principles offer a widely recognized framework emphasizing transparency, accountability, and human-centered values. Their strength lies in broad international adoption and their influence on national strategies. The UNESCO Recommendation on the Ethics of AI provides a comprehensive normative framework, particularly by integrating human rights and social impact considerations into governance discussions. At the regulatory level, the EU AI Act represents a concrete attempt to operationalize risk-based governance. By categorizing AI systems based on risk and imposing corresponding obligations, it offers a practical model for implementation, even though its cross-border implications are still evolving. In terms of technical and organizational practices, standards developed by ISO and IEEE help translate high-level principles into measurable requirements, supporting consistency and auditability. Additionally, impact assessment tools, such as algorithmic impact assessments, are emerging as practical mechanisms to evaluate risks before deployment. These approaches help bridge the gap between principles and operational decision-making. Together, these examples demonstrate that effective AI governance requires a combination of normative frameworks, regulatory measures, and technical standards. The main challenge remains ensuring coherence across these layers and adapting them to different legal and institutional contexts.