Acted

A successful Global Dialogue on AI Governance should be judged by what it actually changes or will potentially change in future by shifting from general agreement to focused, actionable coordination, while ensuring that a wider range of actors can engage in a meaningful way. First, it needs to move beyond broad principles and agree on a small set of practical priorities. The global conversation on AI is already full of high-level commitments. What's missing is alignment on concrete issues such as basic transparency standards, risk classification, and safeguards for high-impact systems. Second, success would mean making progress on coordination between different regulatory approaches. Right now, AI governance is fragmented across regions, and that creates confusion, especially for developing countries with limited capacity. The Dialogue should at least begin to map out how these systems can work together in practice, even if full alignment is unrealistic. Third, inclusion needs to be meaningful. From my experience working with goverments and civil society organisations, participation alone is not enough. A strong outcome would be practical support mechanisms like technical guidance, capacity building, or funding pathways. It can allow underrepresented actors to actually implement and shape AI governance following the discussions. Fourth, the Dialogue should not end as a one-off event. It needs a clear follow-up structure, with defined next steps and links to ongoing UN processes. Without continuity and follow-up, even the best discussions lose momentum.

• Protection and promotion of human rights
• Interoperability of governance approaches
• Transparency, accountability, and human oversight

These four priorities reflect what actually makes AI governance work in practice, especially in contexts where capacity and resources are limited. - AI capacity-building comes first. In many countries and organisations, the issue is not a lack of frameworks, but a lack of skills, tools, and institutional readiness to apply them. From my professional experience in working with AI related projects, without this foundation, even the best policies remain theoretical. - Transparency, accountability, and human oversight are what turn governance into something real. In environments where enforcement is uneven, simple but clear mechanisms like documentation, explainability, and human control are often more effective than complex rules. They make it possible to track decisions, identify risks, and step in when needed. - Protection and promotion of human rights keeps the focus on people. AI systems can easily reinforce existing inequalities, particularly in more vulnerable settings. Keeping human rights at the centre helps ensure that governance is not just technical, but also fair and grounded in real world impacts. - Interoperability of governance approaches would impact on reducing confusion. Right now, different regions are developing their own AI rules, and for many countries and organisations, navigating these differences is difficult. Some level of alignment can make it easier to adopt and apply governance frameworks without being locked out of global systems.

Several cross-cutting issues are not fully captured. 1) Pace of technological change can be a structural challenge. Governance processes are slow, while AI evolves rapidly. Without more adaptive and iterative approaches, there will always be a gap between policy and practice. 2) The role of the private sector, particularly large technology companies shoud be considered. Much of the real power in AI development is concentrated outside governments. Any meaningful governance approach must address how public institutions engage with, influence, and, where necessary, constrain private actors. 3) Context and geographical difference. Many governance models are being developed in high-capacity environments and may not translate well elsewhere. There is a risk of exporting frameworks that are too complex or misaligned with local realities. Governance needs to be adaptable, not one-size-fits-all.

Governance gaps in AI are already showing up in civil society and development work in Tajikistan. The most immediate challenge is limited capacity. Many organisations simply don't have the technical expertise to assess or use AI properly. This leads to extremes: either AI is avoided altogether, or it's adopted too quickly without understanding the risks. In both cases, the potential value is lost. AI can improve efficiency and reach, helping organisations work with limited resources and deliver services more effectively. If used well, it can strengthen decision-making and expand impact. Another issue is fragmentation. Different regions are developing their own AI rules, but for many countries and organisations, these frameworks are difficult to interpret and apply. This creates confusion and slows down progress, while also increasing the risk of falling out of step with international standards. There are also real concerns around accountability. In practice, many AI tools operate as black boxes. In sectors like development or public services, this can lead to biased decisions, exclusion of vulnerable groups, or outcomes that no one can clearly explain or take responsibility for.

The AI Dialogue can play a crucial role by providing a space where countries, civil society, and other stakeholders can find common ground by encouraging alignment, highlighting shared principles, and signalling minimum standards for responsible AI use (before national rules approves). Right now, AI governance is fragmented, there are separate talks in different regions, inside government institutions, between private sector companies. It moves at different speeds with different approaches which increases risks for creating confusion and barriers for collaboration. It can also give underrepresented actors a real voice. Many countries and organisations lack the technical expertise or institutional capacity to shape AI rules. By including these perspectives, the Dialogue can help ensure that global standards aren't dominated by a few powerful states or companies, but reflect a broader range of needs and realities. Another important role is promoting transparency and accountability. By discussing oversight, human control, and documentation, the Dialogue can help move governance from abstract principles to practical, usable measures especially in areas like development, public services, or humanitarian programmes where AI decisions have real social consequences. Finally, the Dialogue can act as a starting point for ongoing engagement, not just a one single event. With clear follow-up steps, milestones, and links to other UN mechanisms, it can turn conversation into continuous international cooperation rather than isolated statements.

The AI Dialogue can build on existing initiatives while filling gaps that currently limit their reach or impact. Key examples include the EU AI Act, the UNESCO Recommendation on the Ethics of AI. These frameworks provide useful guidance, technical standards, and multilateral coordination, but has gaps related to local realities. The Dialogue should also connect with national AI strategies, regional bodies, and civil society networks that work on technology, human rights, and development. These actors bring practical insights and on-the-ground experience, which are critical for governance that actually works beyond high-resource environments. The added value of the AI Dialogue lies in bringing these fragmented efforts together. The Dialogue event can also give a voice to underrepresented actors, making sure global governance reflects a wider range of priorities and contexts. Another key contribution would be turning principles into practice. Many existing initiatives offer guidance, but few provide actionable steps for implementation, especially in resource-constrained settings (recent shortage of international aid funds).

Each stakeholder can contribute to the AI Dialogue in the following possible way: Governments can provide policy perspectives, share national strategies, and highlight regulatory challenges. Their participation is critical to ensure that dialogue outcomes are grounded in practical governance realities. Civil society organisations can bring insights on social impact, human rights, and local implementation challenges. They ensure that discussions go beyond technical standards and consider how AI affects communities, particularly vulnerable groups. Academia and research institutions can contribute evidence-based analysis, technical expertise, and risk assessment frameworks. The private sector can provide knowledge of AI development, usage in practice/deployment, and operational constraints that will contribute to addressing practical feasibility and responsible innovation. Format and structure should encourage both high-level policy discussion and practical, actionable outcomes. I would recommend: 1) Plenary sessions to establish shared principles and identify key priorities. 2) Thematic working groups for deeper dives into issues like capacity-building, accountability, or human rights impacts. It will allow wider participation from underrepresented actors, ensuring diverse perspectives inform decisions 3) Case-study and scenario exercises to explore real-world challenges and solutions. 4) Clear follow-up mechanisms to capture outcomes, assign responsibilities, and maintain momentum between events.

Civil society organisations from developing countries often lack the resources or technical expertise to engage meaningfully. Without them, global standards risk to miss the realities on the ground and reflecting the priorities of high-capacity states and large tech companies. Local communities and marginalised groups including women, youth, indigenous peoples, and low-income populations are rarely consulted. Yet AI systems often have disproportionate impacts on these groups, from biased algorithms to unequal access to services. Small and medium-sized enterprises (SMEs) and local innovators are also underrepresented. Their perspectives on practical implementation, constraints, and opportunities for responsible innovation are essential for workable governance.

I think that in order to foster meaningful engagement, the AI Dialogue shall focus more on interactive, multi-layered approaches that encourage dialogue, problem-solving, and real-world application. 1. Thematic working groups: Small, focused groups can dive deeply into specific issues such as accountability, capacity-building, or human rights impacts. These groups allow participants to exchange practical experiences and co-develop recommendations, rather than just listen to presentations. 2. Scenario-based exercises and simulations conbined with knowledge co-creation sessions: Using real-world case studies or hypothetical scenarios can help participants explore challenges and trade-offs in a safe environment. The second part of the activity would be workshops where participants jointly design tools, frameworks, or guidelines on further immediate actions.. 3. Interactive digital platforms: Online forums, polls, and collaborative workspaces can extend participation beyond in-person attendees. They allow a wider range of stakeholders including those from underrepresented regions. 4. Multi-stakeholder roundtables: Structured discussions that mix government, private sector, academia, and civil society can break silos and promote cross-sector understanding. Rotating facilitation or "fishbowl" formats ensure that quieter voices are heard.

As far as I know several approaches have demonstrated effectiveness in advancing inclusive AI governance. The OECD AI Principles and UNESCO AI Ethics Framework provide globally applicable foundations, though their implementation must be contextualised to developing country realities (including limited regulatory capacity, infrastructure gaps, and data sovereignty concerns). Effective governance requires multi-stakeholder platforms that meaningfully include civil society, academia, and affected communities. Knowledge exchange mechanisms between developing countries and capacity-building programmes are equally vital to learn from peers who've navigated similar constraints. At the organisational level, clear accountability structures, lifecycle monitoring, and mandatory human oversight of high-risk AI decisions shoud be always a priority. Ultimately, governance frameworks must balance innovation enablement with rights protection - ensuring AI serves development goals rather than worsening existing inequalities. As for approaches, context-sensitive, iterative governance (where rules evolve alongside technology) will rpomote effective AI governance. This can include capacity-building programmes, stakeholder consultations, and mechanisms to integrate feedback from affected communities.