UBS (but representing my personal views/experience)

The first Global Dialogue on AI Governance would be successful if it establishes a credible foundation for inclusive, practical, and sustained international cooperation on both the risks and opportunities of AI—while translating principles into real-world impact on people's daily lives. Key Success Outcomes Participation: All countries—regardless of size, economic strength, or technological maturity—are meaningfully represented. Ensuring that developing nations have a genuine voice in shaping AI norms strengthens legitimacy, reduces fragmentation, and avoids a system dominated by a few technologically advanced states. Concrete Commitments: Success requires moving beyond high-level principles toward tangible agreements. This includes interoperable governance frameworks, shared standards for safe and trustworthy AI grounded in human rights, and clear accountability mechanisms. Critically, these should build on and align with existing legal and regulatory frameworks—such as data protection, consumer rights, and product safety regimes—rather than creating parallel systems. Capacity-building must also be prioritized through funding, technical partnerships, and skills development. Practical Implementation in Daily Life: The Dialogue should demonstrate how AI governance improves everyday outcomes for citizens. This could include integrating AI safeguards into public services like healthcare, education, and social protection; embedding transparency and audit requirements into AI systems already in use; and ensuring individuals have accessible recourse mechanisms when harmed. Leveraging existing institutions—courts, regulators, and standards bodies—will enable faster, more credible implementation than relying solely on new structures or long-term roadmaps. Long-Term Impact Ultimately, success lies in building trust among governments, industry, academia, and civil society. By prioritizing practical action over abstract commitments, and embedding safeguards into systems people interact with daily, the Dialogue can ensure AI benefits are shared equitably while risks are effectively managed—turning global consensus into lived reality.

• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models
• AI capacity-building

From my perspective, all four thematic areas identified are essential and closely connected, and each requires urgent and practical action. Safe, secure and trustworthy AI is the starting point. Without trust, AI adoption will remain limited and uneven. There is a need to focus on real safeguards-risk-based standards, system testing, and ongoing monitoring-to ensure AI systems are reliable, aligned with human rights, and protected against misuse. AI capacity-building is equally critical to avoid widening global inequalities. Many countries still lack the infrastructure, skills, and institutional readiness to participate meaningfully in the AI ecosystem. Practical support-such as training, access to compute, and knowledge-sharing partnerships-will be key to ensuring more balanced global participation. Transparency, accountability, and human oversight must be embedded into how AI is actually used in everyday contexts. This means ensuring systems can be audited, decisions can be explained where needed, and clear responsibility is defined when things go wrong. Importantly, these mechanisms should build on existing regulatory frameworks and institutions, rather than creating entirely new layers that are harder to implement. Finally, open-source software, open data, and open AI models can play a powerful role in accelerating innovation and inclusion. When responsibly managed, openness lowers barriers to entry, supports capacity-building, and enables collaboration across borders. At the same time, appropriate safeguards are necessary to prevent misuse. Overall, the priority should be to move beyond high-level principles and ensure these areas translate into practical outcomes that people can experience in their daily lives-building trust, expanding access, and ensuring AI works for everyone.

Yes-while the listed themes are strong and relevant, there are a few cross-cutting issues that feel equally important and sometimes overlooked. One of the biggest gaps is how all of this actually works in practice. It's one thing to agree on principles, but another to make them real in people's daily lives-whether in healthcare, education, finance, or public services. There needs to be more focus on how existing institutions like regulators, courts, and public agencies can realistically implement and enforce these safeguards. Another important area is the economic and workforce impact of AI. Beyond capacity-building, there are real questions around job displacement, reskilling at scale, and how the benefits of AI are distributed. If not addressed early, this could widen inequalities and reduce public trust. Access to infrastructure and compute is also becoming a defining issue. Countries and smaller actors without access to data, computing power, or technical resources risk being left behind. Ensuring more equitable access-through shared infrastructure or partnerships-will be key to meaningful global participation. There is also a growing concern around environmental sustainability. AI systems can be resource-intensive, and their energy use is increasing quickly. This needs to be part of the conversation, especially as countries balance innovation with climate commitments. Finally, there is the risk of fragmentation across different regulatory approaches. We cannot have EU vs Asia vs US taking different stances....as we will leave in a global village. Greater coordination and alignment will be important to avoid a patchwork of rules that slows innovation and creates confusion. Overall, these issues cut across all the themes and are critical to making AI governance not just well-intentioned, but practical, fair, and sustainable.

From a financial services perspective—particularly as a bank headquartered in Switzerland and operating across multiple jurisdictions—governance gaps in AI present both significant challenges and opportunities. One of the main challenges is regulatory fragmentation. Different regions are moving at varying speeds and with different approaches to AI governance. For a global bank, this creates complexity in ensuring consistent standards for safe, secure, and trustworthy AI while remaining compliant with local regulations. The lack of interoperability across frameworks increases operational burden and can slow down responsible innovation. Another key challenge is embedding transparency, accountability, and human oversight into complex financial systems. Many AI use cases—such as credit decisioning, fraud detection, and risk modeling—require a careful balance between performance and explainability. Ensuring auditability, clear accountability, and effective human control, while maintaining efficiency, is an ongoing and resource-intensive effort. Capacity and infrastructure gaps also affect the sector. While large institutions may have access to advanced capabilities, smaller market participants and emerging markets risk being left behind. This can create uneven playing fields and limit broader ecosystem resilience. At the same time, there are important opportunities. Strong AI governance—particularly around safety, transparency, and oversight—can enhance trust with clients, regulators, and society, which is critical in financial services. It also creates a foundation for scaling AI responsibly across markets. In addition, open-source tools and shared standards offer opportunities to accelerate innovation, improve risk management, and reduce duplication of effort—provided they are implemented with appropriate safeguards. Overall, addressing governance gaps in a coordinated and practical way will be key to enabling financial institutions to innovate responsibly while maintaining stability, trust, and compliance across global markets.

The AI Dialogue has the potential to play a very practical and important role in bringing global AI governance closer to real-world implementation. At its core, it can help create greater alignment across countries and sectors. Today, there is a growing risk of fragmented approaches, which is particularly challenging for global industries like financial services. A platform that encourages interoperability and shared understanding can make it much easier to apply consistent standards for safe, secure, and trustworthy AI across markets. Equally important, the Dialogue can help move from discussion to action. Beyond high-level principles, there is a real need for practical guidance—what transparency, accountability, and human oversight look like in day-to-day operations, and how these can be embedded into existing systems, processes, and regulatory frameworks. It also has a key role to play in capacity-building and inclusion, by facilitating partnerships and knowledge-sharing so that more countries can actively participate in and benefit from AI developments. From my perspective as a seasoned financial services professional working across jurisdictions, there is a strong opportunity to contribute more actively to this process. Financial institutions are already dealing with many of these challenges in practice—balancing innovation with risk management, compliance, and client trust—and can offer valuable, experience-based insights into what works on the ground. Overall, the Dialogue can serve as a space not just for alignment, but for practical collaboration, where policymakers and practitioners come together to shape solutions that are both globally relevant and operationally feasible.

The AI Dialogue doesn't need to start from scratch—there is already a strong foundation of initiatives and partnerships it can build on. The real opportunity is to connect these efforts in a more practical and coherent way. At the global level, organizations like the United Nations and the OECD have set important principles and policy directions. Initiatives such as the Global Partnership on AI and the G7 Hiroshima AI Process are also driving collaboration between countries. In parallel, regional frameworks like the EU AI Act are beginning to translate these ideas into more concrete, risk-based rules. In financial services, there are already well-established global coordination mechanisms. Institutions such as the Financial Stability Board, the Bank for International Settlements, and the Basel Committee on Banking Supervision have long experience in aligning standards across jurisdictions, managing risk, and ensuring oversight—many of which are directly relevant to AI. The issue today is less about a lack of activity, and more about how fragmented these efforts can feel in practice. This is where the AI Dialogue can add real value. It can act as a bridge between these different initiatives, helping to align approaches, reduce duplication, and create more consistency across regions and sectors. Just as importantly, it can focus on what implementation actually looks like—turning principles into guidance that can be applied in day-to-day operations. From a practitioner's perspective, having a space where policymakers and industry can engage more directly would be particularly valuable. Ultimately, the Dialogue can help move from parallel conversations to genuine coordination and practical progress.

Different stakeholders can bring real value to the AI Dialogue, but only if it's designed to be practical and collaborative—not just another forum for high-level statements. Governments have an important role in setting direction and aligning regulatory approaches. But they shouldn't do this in isolation. Industry—especially sectors like financial services—can offer hands-on experience from operating in highly regulated, risk-sensitive environments across multiple jurisdictions. Banks are already dealing with questions around model risk, explainability, auditability, and human oversight on a daily basis, and can provide concrete insights into what is actually workable in practice. Academia and technical experts can ensure the Dialogue is grounded in evidence and forward-looking thinking, while civil society brings in the broader societal perspective and helps keep the focus on real-world impact and fairness. In terms of format, it would help to keep things focused and outcome-driven. Smaller working groups around specific themes—such as safe and trustworthy AI or transparency—would likely be more effective than large, general discussions, especially if they are expected to produce tangible outputs like guidance, use cases, or tools. It would also be particularly valuable to create space for direct exchange between policymakers and practitioners, including sector-specific sessions. For example, financial services could share how existing governance frameworks—such as risk management, compliance, and audit structures—can be leveraged for AI, rather than building entirely new systems. Finally, the Dialogue shouldn't be a one-off event. AI is evolving quickly, and there needs to be continuity—regular engagement, shared learning, and the flexibility to adapt. At its best, the AI Dialogue can feel less like a conference and more like a working session, where stakeholders collaborate to solve real problems and deliver outcomes that are both globally relevant and operationally realistic.

There are still several important voices missing—or not heard enough—in global discussions on AI governance. One of the biggest gaps is developing countries and smaller economies. Many simply don't have the resources or access to fully participate, yet they are often the most affected by global decisions. Giving them a more active role—through funding, capacity-building, and real decision-making power—would make a big difference. Another group that's often underrepresented is practitioners from regulated industries, like financial services, healthcare, or the public sector. These are the people already dealing with AI in real-world settings, under strict rules around risk, compliance, and oversight. Their practical experience could really help ground the conversation in what actually works. There's also a need to better include civil society and everyday users, especially those from vulnerable communities. They're directly impacted by AI, but often not involved in shaping it. Making participation more accessible and structured would help bring these perspectives in. Finally, startups and smaller companies are often overlooked, despite being key drivers of innovation.

To foster meaningful and dynamic engagement, the AI Dialogue should move beyond traditional panels and adopt more interactive, problem-solving formats. Small, focused working sessions can be highly effective—bringing together policymakers, industry practitioners, and technical experts to work through real use cases (e.g. AI in finance or healthcare) and co-develop practical solutions. Scenario-based simulations or "stress tests" could also add value, allowing participants to explore how governance frameworks hold up in real-world situations, such as model failures or cross-border risks. Public-private roundtables should be structured as two-way exchanges, not presentations—especially to capture insights from sectors like financial services that are already implementing AI under strict oversight. Finally, pilot initiatives/sandboxes linked to the Dialogue could help test ideas in practice between sessions.

Several existing approaches offer practical lessons for effective AI governance. The EU AI Act provides a clear risk-based framework, linking obligations to the level of risk, which is particularly useful for real-world implementation. It is closely connected to the General Data Protection Regulation, which already sets strong standards for data protection, privacy, and individual rights-together creating a more comprehensive foundation for trustworthy AI. In financial services, established model risk management and audit frameworks-aligned with guidance from bodies like the Basel Committee on Banking Supervision-offer proven approaches to embedding oversight, validation, and accountability into complex systems. On the technical side, model documentation practices such as "model cards" and robust data governance frameworks improve transparency and traceability. Regulatory sandboxes, used in countries like the UK and Singapore, allow AI solutions to be tested in controlled environments while maintaining oversight. Finally, open-source platforms and shared datasets support collaboration and capacity-building, when paired with appropriate safeguards.