Independent Researcher / Student (AI & Machine Learning)

A successful first Global Dialogue on AI Governance would produce clear, actionable, and inclusive outcomes rather than broad principles alone. First, it should establish a shared baseline framework for AI governance that outlines common principles such as transparency, accountability, safety, and fairness, while allowing flexibility for regional adaptation. This would help align efforts across countries without imposing a one-size-fits-all approach. Second, the dialogue should result in practical mechanisms for international collaboration, such as open repositories of best practices, standardized evaluation benchmarks, and shared safety protocols for advanced AI systems. These resources should be accessible to both developed and developing countries. Third, success would include a strong commitment to inclusion of underrepresented regions and low-resource languages. Dedicated initiatives, funding programs, and partnerships should be announced to support AI development in regions like South Asia and Africa, ensuring that global AI systems are more representative and equitable. Fourth, the dialogue should define guidelines for responsible open-source AI, balancing innovation with risk mitigation. This includes clarity on model release strategies, documentation standards, and responsible usage policies. Fifth, establishing capacity-building programs would be a key outcome. This includes investments in education, compute infrastructure, and research opportunities so that emerging economies can actively contribute to AI development. Finally, a successful dialogue should create a clear roadmap with measurable milestones leading up to future meetings, ensuring continuity, accountability, and tangible progress. Overall, success would mean moving from discussion to implementation, with concrete commitments that make AI governance more inclusive, practical, and globally coordinated.

• AI capacity-building
• Open-source software, open data and open AI models
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

My selected priorities reflect both my technical background in AI/ML and the needs of underrepresented regions like South Asia. AI capacity-building is essential to ensure that developing countries can actively participate in AI innovation rather than remain passive consumers. This includes access to education, compute resources, and research opportunities. Open-source software, open data, and open AI models are critical for democratizing AI development. Open ecosystems enable researchers and developers, especially in low-resource settings, to build, adapt, and improve AI systems. However, openness should be balanced with responsible safeguards. Transparency, accountability, and human oversight are fundamental to building trust in AI systems. Clear documentation, explainability, and mechanisms for human control are necessary to ensure that AI systems are used responsibly and ethically. Finally, the social, economic, ethical, cultural, linguistic, and technical implications of AI are particularly important from my perspective working with low-resource languages. Many AI systems underrepresent diverse linguistic and cultural contexts, which can lead to biased or exclusionary outcomes. Addressing these implications ensures that AI systems are inclusive and globally relevant. Together, these priorities support a vision of AI governance that is inclusive, practical, and innovation-friendly, while ensuring that the benefits of AI are distributed equitably across different regions and communities.

Yes, there are several cross-cutting and emerging issues that are not fully captured by the listed themes. One key issue is data governance and data equity. While AI governance often focuses on models, the quality, ownership, and representativeness of data are equally important. There is a need for frameworks that ensure fair access to datasets, respect for local data sovereignty, and inclusion of underrepresented languages and communities. Another emerging concern is compute inequality. Access to high-performance computing resources is highly concentrated in a few countries and organizations, creating barriers for researchers and startups in developing regions. Addressing this imbalance is critical for equitable participation in AI development. Evaluation and benchmarking standards for AI systems, especially large language models, are also still evolving. There is a lack of globally accepted, transparent, and culturally inclusive evaluation frameworks, particularly for low-resource languages. This makes it difficult to assess safety, fairness, and performance consistently across regions. Additionally, AI alignment with local cultural and societal values is an important but underexplored issue. Governance frameworks should consider that acceptable use, risk perception, and ethical norms may vary across different societies. Finally, the rapid rise of multimodal and general-purpose AI systems introduces new governance challenges, as these systems can be applied across domains with varying levels of risk. This calls for more adaptive and context-aware regulatory approaches. Addressing these cross-cutting issues will be essential to ensure that AI governance is not only technically robust but also globally inclusive and future-ready.

In Nepal and the broader South Asian region, governance gaps in AI are creating both significant challenges and emerging opportunities. One major challenge is limited AI capacity and infrastructure. Access to compute resources, high-quality datasets, and advanced training opportunities remains constrained. This limits the ability of local researchers and startups to develop competitive AI systems and contributes to dependence on external technologies. Another critical issue is the underrepresentation of low-resource languages such as Nepali in AI models. Many widely used systems perform poorly in local languages, leading to unequal access to AI benefits and potential cultural and linguistic biases. The lack of standardized datasets and evaluation benchmarks further exacerbates this gap. There are also gaps in transparency and accountability frameworks. As AI adoption grows across sectors like education, finance, and public services, the absence of clear guidelines on responsible use, data privacy, and model explainability increases the risk of misuse and erodes trust. However, these challenges also present important opportunities. The growing global focus on open-source AI and collaborative development enables researchers in Nepal to contribute to and benefit from shared models, datasets, and tools. This lowers barriers to entry and accelerates innovation. Additionally, increased international attention on inclusive AI governance creates opportunities for countries like Nepal to actively shape policies that reflect local needs and values, rather than adopting external frameworks uncritically. Finally, investments in AI capacity-building and education have the potential to empower a new generation of developers and researchers, enabling the region to move from being technology consumers to active contributors. Addressing these gaps can help ensure that AI development in Nepal is inclusive, locally relevant, and globally connected.

The AI Dialogue can play a critical role as a neutral, inclusive, and action-oriented platform for advancing international cooperation on AI governance. First, it can help bridge gaps between countries at different stages of AI development by ensuring that developing nations have an equal voice in shaping global norms. This is essential to avoid governance frameworks that are dominated by a small number of technologically advanced regions. Second, the Dialogue can facilitate knowledge sharing and coordination, including best practices, policy approaches, safety standards, and evaluation methods. Creating open, accessible repositories and collaborative working groups would enable stakeholders across governments, academia, and industry to learn from each other. Third, it can promote interoperability of governance approaches, helping align national and regional policies while respecting local contexts. This reduces fragmentation and supports the development of globally compatible AI systems and regulations. Fourth, the Dialogue can act as a catalyst for joint initiatives, such as multinational research collaborations, shared datasets, and capacity-building programs. These efforts can lower barriers to participation for underrepresented regions and accelerate inclusive innovation. Fifth, it can support the development of guidelines for responsible open-source AI and emerging technologies, ensuring a balanced approach between innovation and risk mitigation. Finally, the AI Dialogue can provide continuity and accountability by establishing clear roadmaps, measurable goals, and follow-up mechanisms between annual meetings. Overall, its role should be to move beyond discussion and actively enable coordination, inclusion, and implementation, ensuring that AI governance evolves as a truly global and cooperative effort.

The AI Dialogue should build upon and connect with existing global and multi-stakeholder initiatives to avoid duplication and accelerate progress. Key initiatives include the OECD AI Principles, which provide widely adopted guidelines on trustworthy AI; the Global Partnership on AI (GPAI), which advances research and collaboration on responsible AI; and UNESCO Recommendation on the Ethics of Artificial Intelligence, which emphasizes human rights and ethical governance. In addition, emerging efforts such as the AI Safety Institutes in multiple countries and open-source collaborations like Hugging Face play an important role in shaping technical standards and accessibility. The AI Dialogue can add value by acting as a unifying and coordinating platform that connects these efforts across regions and sectors. Unlike existing initiatives that may be regional or domain-specific, the Dialogue can provide a truly global and inclusive forum, ensuring participation from underrepresented countries and communities. It can also enhance interoperability between frameworks, helping align principles, standards, and policy approaches to reduce fragmentation in AI governance. This is particularly important as different countries adopt varying regulatory models. Furthermore, the Dialogue can promote practical implementation by translating high-level principles into actionable guidelines, toolkits, and shared resources. It can also support capacity-building partnerships, linking developing countries with established institutions for knowledge transfer, training, and infrastructure support. Finally, by fostering continuous engagement and accountability, the AI Dialogue can ensure that existing initiatives evolve in a coordinated manner and remain responsive to emerging challenges. Overall, its added value lies in connecting fragmented efforts, amplifying inclusivity, and driving coordinated global action on AI governance.

Different stakeholders can contribute to the AI Dialogue by bringing complementary expertise, perspectives, and resources. Governments can provide regulatory insights, national strategies, and policy frameworks, while committing to interoperable and inclusive governance approaches. The private sector can share technical expertise, real-world deployment experience, and best practices for safety, transparency, and responsible innovation. Academia and the technical community can contribute research on AI safety, evaluation, and emerging risks, as well as develop open benchmarks and tools. Civil society organizations play a critical role in representing public interests, advocating for human rights, and highlighting social and ethical impacts. International organizations can facilitate coordination, provide neutral platforms, and support capacity-building across regions. To be effective, the AI Dialogue should adopt a multi-layered and action-oriented structure. First, it should include thematic working groups (e.g., safety, open-source AI, capacity-building) that operate continuously between annual meetings and produce concrete outputs. Second, the Dialogue should combine high-level plenary sessions with technical workshops and roundtables, enabling both policy alignment and in-depth technical discussions. Third, it should support open consultation mechanisms, allowing broader stakeholders including those from underrepresented regions to submit inputs and feedback. Fourth, the Dialogue should produce tangible deliverables, such as guidelines, toolkits, shared datasets, and evaluation frameworks. Finally, a clear roadmap with milestones and accountability mechanisms should be established to track progress across meetings. Overall, a structured, inclusive, and implementation-focused approach will ensure that the AI Dialogue translates diverse contributions into meaningful global outcomes.

Several important voices remain underrepresented in global discussions on AI governance. First, communities from developing and least-developed countries, particularly in regions like South Asia and Africa, are often excluded from decision-making processes. This leads to governance frameworks that may not reflect their local realities, infrastructure constraints, or societal needs. Second, speakers of low-resource languages are significantly underrepresented. Most AI systems are optimized for a small number of dominant languages, which risks marginalizing linguistic diversity and limiting access to AI benefits. Third, grassroots communities, local innovators, and small startups often lack the resources or platforms to contribute, despite being directly affected by AI deployment in areas such as education, agriculture, and public services. Fourth, interdisciplinary perspectives, including social scientists, ethicists, and cultural experts from diverse backgrounds, are sometimes overlooked in favor of purely technical or policy-driven viewpoints. To address these gaps, the AI Dialogue should prioritize inclusive participation mechanisms. This includes providing financial support, travel grants, and remote participation options for stakeholders from underrepresented regions. It should also promote multilingual engagement, including translation of materials and support for non-English contributions. Establishing regional consultation forums and partnerships with local institutions can help bring grassroots perspectives into global discussions. Additionally, supporting open platforms and community-driven initiatives will enable broader participation from independent researchers and developers. By actively including these voices, the AI Dialogue can ensure that AI governance is more equitable, representative, and responsive to global diversity.

To foster meaningful and dynamic engagement, the AI Dialogue could adopt a mix of interactive, inclusive, and technology-enabled formats. First, thematic breakout sessions and workshops allow stakeholders to dive deeply into specific topics such as AI safety, open-source models, or low-resource language inclusion. These smaller, focused groups encourage detailed discussion, knowledge sharing, and practical recommendations. Second, multistakeholder roundtables bring together government representatives, private sector experts, civil society, academia, and technical communities to debate policy, ethics, and implementation strategies. Rotating moderators and structured dialogue prompts can ensure equitable participation. Third, virtual and hybrid participation platforms can expand accessibility for stakeholders from underrepresented regions, enabling remote attendance, live polls, Q&A, and collaborative document editing in real time. Fourth, hands-on technical demos and sandbox sessions can showcase AI models, datasets, and evaluation tools. This allows policymakers and non-technical participants to understand capabilities, limitations, and risks, promoting informed decision-making. Fifth, challenge-based or hackathon-style events could stimulate innovation while engaging participants in collaborative problem-solving, highlighting practical solutions to pressing AI governance issues. Sixth, multilingual discussion channels and translation support ensure broader inclusivity, particularly for low-resource language communities often excluded from global dialogues. Finally, post-session synthesis and collaborative reporting with open-access outputs can capture insights, actionable recommendations, and emerging trends, making the Dialogue's outcomes tangible and widely shareable. By combining interactive, inclusive, and results-oriented formats, the AI Dialogue can move beyond presentations to active engagement, creating a vibrant environment where diverse perspectives contribute to actionable, globally relevant AI governance strategies.

Several policies, practices, platforms, and approaches have demonstrated effective AI governance and offer concrete solutions to current challenges. Policies and Guidelines: The OECD AI Principles and UNESCO Recommendation on the Ethics of Artificial Intelligence provide globally recognized frameworks for trustworthy AI, emphasizing transparency, accountability, safety, and human rights. These guidelines help governments and organizations establish baseline governance structures. Regulatory Practices: The European Union's AI Act exemplifies risk-based regulation, offering a scalable model for governing high-impact AI systems while allowing innovation in lower-risk applications. Platforms and Collaborative Initiatives: Open-source ecosystems like Hugging Face and partnerships such as the Global Partnership on AI enable shared model development, dataset access, and cross-border collaboration. These platforms promote transparency, reproducibility, and equitable access to AI tools. Capacity-Building and Education: Initiatives like AI4D (Artificial Intelligence for Development) and regional AI research hubs help train local talent, provide compute resources, and support participation in global AI research, addressing capacity gaps in underrepresented regions. Technical Approaches: Implementing explainable AI (XAI), model documentation standards (e.g., datasheets for datasets, model cards), and risk-tiered deployment strategies ensures that AI systems are transparent, accountable, and aligned with human oversight. Human-Centered Approaches: Engaging civil society, ethical review boards, and multidisciplinary advisory committees helps ensure that AI systems consider social, cultural, linguistic, and ethical implications, reducing bias and unintended harm. These combined approaches demonstrate that effective AI governance requires multi-stakeholder engagement, practical implementation mechanisms, open collaboration, and continuous evaluation to address technical, ethical, and social challenges.