National Information Technologies JSC (NITEC)

A successful first Global Dialogue on AI Governance would be one that helps translate global discussions into practical directions that governments can apply in their national contexts. Today, many countries, including Kazakhstan, are actively integrating AI into digital public infrastructure and public service delivery. In this context, it is important that the Dialogue supports a clearer understanding of how AI governance principles can be implemented in practice - particularly in areas such as data governance, interoperability, and accountability within government systems. Another important outcome would be the strengthening of international cooperation. There is significant value in creating space for countries to exchange experiences, align approaches where possible, and explore the reuse of existing digital and AI solutions. This is especially relevant for countries building and scaling digital public infrastructure. Capacity building should also be a key element of success. While interest in AI adoption is growing rapidly, many governments are still developing the necessary technical expertise and institutional frameworks. Practical initiatives focused on knowledge sharing and skills development would be highly beneficial. Finally, it would be important for the Dialogue to serve as a starting point for continued engagement. Establishing a platform or community for ongoing exchange would help ensure that discussions evolve into tangible collaboration and long-term impact. For Kazakhstan, the value of the Dialogue will be measured by its ability to support practical implementation and foster meaningful international partnerships in the field of AI governance.

• Transparency, accountability, and human oversight
• AI capacity-building
• Interoperability of governance approaches
• Open-source software, open data and open AI models

Our selection reflects areas where we are already facing practical questions while deploying AI within government systems. Capacity-building remains one of the main constraints. While interest in AI is growing quickly, the ability to design, implement and manage such systems within government is still limited. This applies not only to technical skills, but also to data governance, product thinking and understanding how AI should be used in public services. Interoperability is equally important. In Kazakhstan, AI is being introduced on top of existing digital infrastructure, including data platforms and integration layers. This makes it essential to ensure that governance approaches are aligned and can work across systems, rather than being developed in isolation. At the same time, as AI becomes part of decision-making processes, questions around transparency, accountability and human oversight become very practical. These are not abstract principles - they directly affect how trust is maintained in public services. We also see strong potential in open-source approaches and the use of open data and models. For countries building digital ecosystems, this is often the most realistic way to accelerate development, avoid duplication and enable collaboration. Overall, these priorities reflect a focus on implementation and day-to-day realities of introducing AI into government systems.

One important cross-cutting issue is the integration of AI governance into existing digital government infrastructure. In practice, AI is rarely deployed as a standalone solution - it is embedded into data platforms, registries, and service delivery systems. This creates challenges that are not fully reflected in current thematic areas, particularly around how governance mechanisms are applied within complex, interconnected systems. Another emerging issue is the operationalization of governance. While many principles and frameworks already exist, there is still limited clarity on how these should be translated into day-to-day processes - for example, how to implement oversight in automated systems, how to audit AI models in production, or how to assign responsibility across institutions. This gap between policy and implementation remains significant. A related point is the sustainability of AI systems in the public sector. Governments need to consider not only development, but also long-term maintenance, updates, infrastructure costs, and dependency on external vendors or technologies. This is particularly relevant for countries building national AI capabilities. Finally, there is a growing need to better align AI governance with broader DPI strategies. AI does not exist in isolation - its effectiveness depends on data availability, interoperability, and institutional coordination. Addressing these interdependencies is essential for achieving meaningful impact.

In Kazakhstan, the rapid development of digital public infrastructure has created a strong foundation for AI adoption. However, the pace of technological implementation is currently ahead of governance frameworks, which creates a number of practical challenges. One of the main gaps relates to capacity. While AI solutions are being introduced into public services, there is still limited experience within government institutions in managing the full lifecycle of these systems - from design and procurement to monitoring and evaluation. This affects both the quality of implementation and the ability to scale solutions across sectors. Interoperability also remains a challenge at the governance level. Although technical integration between systems is well developed, approaches to data access, model deployment, and responsibility for AI-driven decisions are not always aligned across institutions. This can slow down implementation and create uncertainty in cross-agency use cases. At the same time, the introduction of AI into government processes increases the importance of transparency and accountability. In practice, this raises questions around explainability, auditability, and the role of human oversight, particularly in services that directly affect citizens. There are also opportunities. Kazakhstan's existing digital infrastructure, including national data platforms and integration systems, creates favorable conditions for scaling AI solutions in a structured way. In addition, growing international cooperation, including with development partners - provides access to expertise, standards, and financing that can support more robust governance approaches. Overall, current gaps are not a barrier, but rather a transition point - creating an opportunity to embed governance mechanisms directly into the next stage of digital and AI development.

The AI Dialogue can play a practical role by becoming a working platform, not just a discussion format. One of its key functions could be enabling direct exchange between governments that are already implementing AI. Instead of only sharing high-level approaches, countries should be able to present concrete use cases, challenges, and solutions, and receive targeted feedback from peers facing similar issues. The Dialogue could also support more structured collaboration. For example, it can facilitate joint pilots, where several countries test similar approaches to AI governance or deployment in parallel and share results. This would help move faster from theory to practice. Another important role is helping countries access and reuse existing solutions. Many governments are facing similar challenges, and there is clear value in adapting proven tools, frameworks, and open-source components rather than developing everything independently. The Dialogue can act as a connector, linking countries to these resources and to partners who can support implementation. It can also help match countries with different levels of experience - those with more advanced systems can support others through peer-to-peer exchange, mentoring, and technical cooperation. Finally, the Dialogue should provide continuity by maintaining a network of practitioners who stay engaged beyond the event itself. This would allow governments to come back with new challenges, share progress, and build long-term cooperation. In this way, the Dialogue can become a space where countries not only discuss AI governance, but actively help each other implement it.

The AI Dialogue can build on a number of existing initiatives, but its added value would be in connecting them more closely to implementation. There are already strong global efforts shaping AI governance - including OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, and ongoing work within the UN system. At the same time, initiatives such as Digital Public Goods Alliance and the 50in5 campaign are supporting countries in building DPI and adopting reusable digital solutions. Development partners, including the World Bank and regional development banks, are also actively supporting AI and GovTech programs. However, these efforts are often fragmented and operate in parallel. The AI Dialogue can help bridge this gap by creating a space where governance frameworks, technical solutions, and financing mechanisms are discussed together, rather than separately. One important added value would be linking principles to practice. While many frameworks already define "what" responsible AI should look like, there is still limited guidance on "how" to implement this within government systems. The Dialogue can facilitate exchange on practical approaches, tools, and institutional models. Another contribution could be better coordination between countries and partners. The Dialogue can help align priorities, reduce duplication of efforts, and support more targeted collaboration, including joint projects and shared use of digital solutions. Finally, it can strengthen connections between policy and implementation communities. Bringing together policymakers, technical teams, and operators of digital platforms would help ensure that AI governance is both realistic and applicable in practice. In this way, the AI Dialogue can act as a bridge - not replacing existing initiatives, but making them more connected, practical, and usable for governments.

From Kazakhstan's perspective, the added value of the AI Dialogue lies in connecting discussions on governance with the realities of implementation. In practice, different stakeholders are often engaged at different stages: policymakers define principles, technical teams build systems, and development partners support programs. However, these processes are not always aligned. One of the key roles of the Dialogue could be to bring these perspectives together around concrete challenges that emerge when AI is deployed at scale. Governments can contribute not only use cases, but also a clearer understanding of where existing governance approaches are difficult to apply in integrated systems. This is particularly relevant in environments where AI is built on top of existing digital infrastructure and data platforms. International organizations and development partners can help translate these challenges into structured support, including methodologies, capacity-building, and implementation programs. The private sector and research community can contribute by testing approaches in practice and adapting solutions to public sector contexts. In terms of structure, it would be valuable to move beyond general discussions and include sessions focused on specific implementation challenges, as well as space to explore joint approaches or pilots across countries. A Dialogue structured in this way can help shift cooperation from exchange of views to more coordinated and practical action.

Global discussions on AI governance have become more inclusive over time. At the same time, some perspectives could be more consistently reflected. One such perspective is that of practitioners within government - teams directly involved in designing, deploying, and operating AI systems. As AI becomes embedded in public services, many governance challenges emerge at the implementation stage, particularly in the context of existing digital infrastructure. Reflecting this experience more systematically would make discussions more practical and grounded. There is also value in ensuring stronger participation from countries that are actively developing and scaling DPI. These countries are increasingly contributing to AI adoption in government and can offer relevant experience in areas such as interoperability, data use, and service delivery at scale. In addition, involving multidisciplinary teams including technical, data, and product perspectives - can help bridge the gap between policy and implementation.

Case-based sessions could be particularly effective, where countries present specific use cases and implementation challenges, and receive targeted input from peers. This helps move discussions closer to practical solutions. Smaller, focused working groups can also create space for more open and detailed exchange - especially on topics such as data governance, interoperability, or deployment of AI in public services. At the same time, there is strong value in building on existing platforms and communities that already support countries in implementing digital solutions. Initiatives such as 50in5 demonstrate how peer learning, reuse of digital components, and coordinated support can accelerate progress. The Dialogue could connect with such platforms and use them as a foundation for continued collaboration. It would also be useful to include formats that enable ongoing support - for example, matching countries with similar priorities, or linking them with partners who can support implementation. Finally, ensuring follow-up will be important. Sessions could identify concrete next steps, such as peer exchange, joint pilots, or continued engagement through existing communities.

Effective AI governance is increasingly shaped not only by standalone policies, but by how governance is embedded into digital systems and institutional practices. One practical approach is the integration of AI into national digital infrastructure, rather than treating it as a separate layer. In Kazakhstan, AI solutions are developed on top of existing data platforms and service delivery systems, which allows governance requirements - such as data access control, traceability, and oversight - to be applied consistently. This reduces fragmentation and makes governance more enforceable in practice. Another important direction is the shift from static regulation to "built-in" governance. Instead of relying only on external guidelines, governance mechanisms are increasingly embedded into system design - for example, through logging of automated decisions, auditability of models, and clear points for human intervention in high-impact services. There is also growing recognition of the role of reusable digital components and open approaches. Platforms such as GovStack and initiatives under the Digital Public Goods ecosystem demonstrate how shared building blocks can support both faster deployment and greater transparency. For many countries, this also helps reduce dependency on proprietary solutions and supports more sustainable development.