GIZ India

A successful first Global Dialogue on AI Governance would be defined less by declarations and more by tangible convergence, credibility, and continuity. First, convergence on core principles would be critical. While complete consensus is unrealistic, agreement on a minimum set of shared norms - such as transparency, accountability, safety, and human rights - would signal meaningful progress. Importantly, these principles should reflect both Global North and Global South priorities, avoiding a one-size-fits-all approach. Second, the Dialogue should produce actionable pathways, not just high-level commitments. This could include the creation of working groups on priority themes (e.g., risk classification, data governance, AI in public services), timelines for deliverables, and mechanisms for technical cooperation. A roadmap for interoperable regulatory approaches would be especially valuable. Third, inclusive representation and voice equity would determine legitimacy. Success would mean that governments, industry, academia, and civil society - especially from developing countries - have shaped the agenda and outcomes, rather than being passive participants. Fourth, the Dialogue should catalyze capacity-building commitments, particularly for countries with emerging AI ecosystems. This includes knowledge-sharing platforms, funding support, and institutional partnerships to bridge governance gaps. Fifth, trust-building among stakeholders is essential. Even informal alignment between major AI powers on risk mitigation and responsible deployment would be a significant achievement, reducing fragmentation and regulatory arbitrage. Finally, success would depend on institutional continuity. Establishing a formal mechanism - such as an annual forum, a permanent secretariat, or integration with existing multilateral platforms - would ensure that the Dialogue is not a one-off event but the beginning of a sustained global process. In essence, success lies in moving from fragmented conversations to a structured, inclusive, and action-oriented global governance ecosystem for AI.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Open-source software, open data and open AI models

My selection prioritizes outcomes that move beyond symbolic dialogue to practical, inclusive, and sustained global coordination, which is essential in a fast-evolving domain like AI. First, I emphasize convergence on core principles because AI governance is currently fragmented across jurisdictions. Without a shared baseline, regulatory divergence can create loopholes, increase compliance burdens, and undermine safety standards. Even partial alignment helps set a common direction while allowing contextual flexibility. Second, the focus on actionable pathways reflects a key gap in many global forums - strong intent but weak follow-through. By proposing working groups, timelines, and technical cooperation, the Dialogue can translate discussion into implementable outcomes, ensuring relevance for policymakers and practitioners alike. Third, inclusive representation is central to legitimacy. AI systems are global in impact but uneven in development and deployment. Ensuring meaningful participation from the Global South helps surface diverse risks (e.g., bias, exclusion, capacity constraints) and prevents governance frameworks from being dominated by a few advanced economies. Fourth, capacity-building commitments are necessary to operationalize governance. Many countries lack the technical, regulatory, and institutional capabilities to implement AI safeguards. Without targeted support, global standards risk deepening digital divides rather than mitigating them. Fifth, trust-building is a pragmatic priority. Given geopolitical competition in AI, even limited cooperation on safety and risk mitigation can reduce harmful externalities and foster stability. Finally, institutional continuity ensures that momentum is not lost. A one-time dialogue cannot keep pace with AI advancements; sustained platforms are needed to adapt norms, share learning, and monitor progress over time. Together, these criteria reflect a balance between norm-setting, implementation, equity, and long-term governance, which I see as essential for meaningful global AI coordination.

Yes, while most global AI governance discussions cover ethics, safety, and regulation, several cross-cutting and emerging issues often remain under-emphasized: 1. Compute and infrastructure inequality Access to advanced compute power and data infrastructure is becoming a key determinant of AI capability. This creates structural dependency of the Global South on a few firms and countries, raising concerns around digital sovereignty and equitable participation in AI development. 2. Labour market transitions and "invisible work" Beyond job displacement, AI is reshaping work through gig-based data labeling, content moderation, and algorithmic management. These forms of "hidden labour" often lack protections and are concentrated in developing countries, making this both a labour rights and global equity issue. 3. Environmental sustainability of AI The energy and water footprint of large AI systems is significant but insufficiently governed. As countries pursue AI adoption, the tension between digital growth and climate commitments will intensify, especially for resource-constrained regions. 4. Public sector AI and state capacity AI use in governance (e.g., welfare targeting, policing, service delivery) raises unique accountability risks. Many governments lack the institutional capacity to procure, audit, and oversee AI systems effectively, which can amplify exclusion or bias at scale. 5. Data governance beyond privacy Current debates often center on personal data protection, but issues like non-personal data access, community data rights, and data value-sharing are gaining importance-particularly for sectors like agriculture, health, and urban systems. 6. Geopolitics and standards-setting power AI governance is increasingly shaped by strategic competition, influencing standards, supply chains, and alliances. This risks fragmenting the global governance landscape and limiting policy autonomy for smaller economies. Addressing these cross-cutting issues is essential to ensure that AI governance is not only safe and ethical, but also equitable, sustainable, and globally representative.

In the Indian context, governance gaps in emerging AI domains are already shaping both development trajectories and risk exposure - particularly across public systems, labour markets, and digital infrastructure. A key challenge is capacity asymmetry in public sector AI adoption. While AI is increasingly used in welfare delivery, agriculture advisories, and urban governance, institutional capacity for procurement, auditing, and oversight remains uneven across states. This creates risks of exclusion (e.g., errors in beneficiary targeting), limited accountability in algorithmic decision-making, and over-reliance on private vendors. Second, data governance remains incomplete beyond privacy. While India has made progress on personal data protection, frameworks for non-personal data, data sharing, and community data rights are still evolving. This affects sectors like health, skilling, and agriculture, where high-quality, accessible datasets are critical for building inclusive AI solutions. Third, labour market disruptions and informalisation are significant. AI is accelerating shifts toward platform-based and task-based work (e.g., data annotation, gig services), often without adequate social protection. At the same time, there is a gap in aligning skilling systems with emerging AI-linked roles, particularly for women and rural youth. Fourth, compute and infrastructure dependency poses a strategic constraint. Limited domestic access to high-end compute infrastructure increases reliance on global technology providers, raising concerns around affordability, innovation capacity, and digital sovereignty. However, these gaps also present opportunities. India can leapfrog through public digital infrastructure (e.g., digital identity, data exchange frameworks) to build scalable, inclusive AI applications. There is also strong potential to position itself as a leader in responsible and frugal AI innovation, tailored to low-resource settings. Finally, targeted investments in AI skilling, governance capacity, and open data ecosystems can enable India not only to manage risks but to shape global norms - especially by foregrounding equity, inclusion, and development-oriented AI use cases.

The AI Dialogue can serve as a bridging platform that translates fragmented national approaches into more coordinated, inclusive, and action-oriented global governance. First, it can enable norm convergence by facilitating agreement on a core set of interoperable principles and risk frameworks. Rather than pushing for uniform regulation, the Dialogue can help align approaches (e.g., risk classification, safety standards), reducing regulatory fragmentation while respecting national contexts. Second, it can act as a platform for technical cooperation and knowledge exchange. Countries are at very different stages of AI readiness; structured peer learning, sharing of regulatory toolkits, and joint research initiatives can help close capacity gaps - especially for developing economies. Third, the Dialogue can catalyze collective action on global public goods. This includes building shared datasets for development use cases (e.g., climate, health), advancing open and responsible AI models, and coordinating on safety research for frontier systems. Fourth, it can support trust-building and confidence measures among countries and key stakeholders. In a context of geopolitical competition, even limited cooperation - such as information-sharing on AI risks, incident reporting, or safety benchmarks - can reduce uncertainty and prevent harmful outcomes. Fifth, the Dialogue can play a role in mobilizing resources and partnerships. By bringing together governments, industry, and multilateral organizations, it can unlock financing and institutional support for capacity-building, particularly in areas like governance infrastructure, skilling, and compute access. Finally, it can ensure continuity and accountability by establishing follow-up mechanisms - such as working groups, progress tracking, and periodic reviews - so that commitments translate into measurable outcomes. In essence, the AI Dialogue can move global AI governance from isolated efforts to coordinated ecosystems, balancing innovation with safety, and national priorities with collective global interests.

The AI Dialogue should build on existing global and multi-stakeholder efforts to avoid duplication and instead strengthen coherence across the AI governance ecosystem. Key initiatives include the OECD AI Principles and AI Policy Observatory, which provide widely endorsed normative guidance and comparative policy insights; the UNESCO Recommendation on the Ethics of AI, which emphasizes human rights and inclusion; and the Global Partnership on AI, which supports applied research and multi-country collaboration. The Dialogue should also connect with more recent governance platforms such as the G7 Hiroshima AI Process, which focuses on safety and trustworthy AI among advanced economies, and the AI Safety Summit process, which has catalyzed global attention on frontier AI risks. Additionally, regional and development-oriented initiatives - such as digital public infrastructure collaborations and South-South partnerships - are important for grounding governance in real-world implementation contexts. Added value of the AI Dialogue: First, it can act as a convergence layer, bringing together these parallel efforts into a more coordinated framework, especially by bridging Global North–Global South divides that many existing initiatives only partially address. Second, it can provide a more inclusive and development-oriented platform, ensuring that emerging economies shape - not just adopt - global AI norms, with stronger emphasis on equity, capacity-building, and public sector use. Third, it can focus on operationalization, linking high-level principles (e.g., OECD, UNESCO) with practical tools, financing mechanisms, and institutional support for implementation. Finally, the Dialogue can offer continuity across fragmented forums, creating structured follow-up, tracking progress, and fostering sustained cooperation rather than one-off commitments. In sum, its value lies in connecting, contextualizing, and operationalizing existing efforts into a more coherent and globally representative AI governance architecture.

A meaningful AI Dialogue requires clearly defined stakeholder roles and a structured, action-oriented format to translate discussion into outcomes. Stakeholder contributions :- 1. Governments - Provide policy direction, share regulatory experiences, and commit to interoperable frameworks. They should also identify priority areas for cooperation (e.g., safety standards, public sector AI). 2. Industry - Contribute technical expertise, disclose best practices on risk management, and support responsible innovation. Industry can also enable access to tools, compute, and sandboxes for wider ecosystem use. 3. Academia & research institutions - Offer independent evidence, advance safety research, and develop evaluation benchmarks. Their role is critical in bridging technical complexity with policy design. 4. Civil society - Ensure accountability, represent affected communities, and foreground rights-based and equity perspectives especially for marginalized groups. 5. Multilateral organizations - Facilitate coordination, mobilize financing, and anchor the Dialogue within existing global governance frameworks. Recommendations on format and structure :- 1. Thematic working groups - Organize the Dialogue into focused tracks (e.g., safety, data governance, labour, public sector AI), each with clear deliverables and timelines. 2. Multi-stakeholder co-leadership - Each track should be co-led by representatives from different stakeholder groups and regions to ensure balance and ownership. 3. Hybrid structure (plenary + technical tracks) - Plenary sessions for political alignment and agenda-setting Technical tracks for in-depth problem-solving and drafting outputs 4. Action-oriented outputs - Require each track to produce tangible deliverables - toolkits, model guidelines, pilot collaborations, or policy roadmaps. 5. Inclusion mechanisms - Dedicated platforms (e.g., regional consultations, fellowships) to ensure meaningful participation from developing countries and underrepresented groups. 6. Continuity and accountability - Establish a light secretariat, periodic progress reviews, and public reporting to track commitments. This structure ensures the Dialogue is inclusive, technically grounded, and implementation-focused, enabling diverse stakeholders to contribute while driving measurable progress in global AI governance.

Global AI governance discussions are still shaped by a relatively narrow set of actors, leaving several critical voices underrepresented :- 1. Global South policymakers and practitioners Many developing countries are rule-takers rather than rule-shapers. Their contexts informality, low state capacity, linguistic diversity are often insufficiently reflected in global norms. 2. Workers in AI value chains This includes data annotators, content moderators, and gig workers who sustain AI systems but lack visibility, labour protections, and representation in governance debates. 3. Marginalized and vulnerable communities Women, rural populations, linguistic minorities, and persons with disabilities are often discussed as beneficiaries, not decision-makers - despite being disproportionately affected by bias and exclusion in AI systems. 4. Small and medium enterprises (SMEs) and local innovators AI governance is frequently shaped by large technology firms, while smaller players face compliance burdens without having a voice in shaping standards. 5. Public sector implementers at subnational levels Frontline officials (e.g., in districts or municipalities) who deploy AI systems in welfare, health, or agriculture are rarely included, even though they face real operational challenges. How to include them :- 1. Institutionalized representation - Reserve seats or quotas for underrepresented regions and stakeholder groups in working groups and decision-making bodies. 2. Resourced participation - Provide travel grants, fellowships, and translation support to enable meaningful engagement not just symbolic inclusion. 3. Decentralized consultations - Conduct regional and sectoral consultations (especially in the Global South) to feed into global agendas. 4. Participatory mechanisms - Use citizen assemblies, worker forums, and civil society platforms to capture lived experiences and ground-level risks. 5. Capacity-building linkages - Pair participation with training and technical support so stakeholders can engage substantively. Ensuring these voices are included is essential not just for equity, but for designing AI governance that is realistic, context-sensitive, and globally legitimate.

To move beyond passive participation, the AI Dialogue should adopt interactive, problem-solving–oriented formats that balance technical depth with inclusive engagement: 1. Policy Labs ("co-creation sprints") Small, diverse groups work intensively on a defined challenge (e.g., AI risk classification for public services) and produce draft frameworks or toolkits within a fixed time. This shifts the Dialogue from discussion to co-development of solutions. 2. Scenario-based simulations Participants engage in real-world case simulations (e.g., AI failure in welfare targeting or misinformation during elections), taking on roles (government, industry, civil society). This builds shared understanding of trade-offs, risks, and coordination needs. 3. Reverse panels ("voices from the margins") Instead of experts speaking about communities, affected stakeholders as gig workers, frontline officials, or marginalized users lead the discussion, with policymakers responding. This ensures lived experiences directly shape policy thinking. 4. Multi-stakeholder "clinics" Countries or institutions present specific governance challenges (e.g., lack of audit capacity), and a cross-sector panel provides targeted, practical recommendations. This creates peer-learning with immediate applicability. 5. Interactive digital platforms Use real-time polling, collaborative drafting tools, and open comment platforms to crowdsource inputs on principles, guidelines, or draft outputs enabling broader and asynchronous participation. 6. AI sandbox demonstrations Live demonstrations of AI systems, audits, or regulatory sandboxes allow participants to engage with practical tools and governance mechanisms, bridging theory and implementation. 7. Commitment roundtables Stakeholders publicly announce measurable commitments (e.g., funding, partnerships, pilot projects), followed by tracking mechanisms driving accountability and continuity. These formats can make the Dialogue more participatory, grounded, and outcome-driven, ensuring that diverse stakeholders not only contribute ideas but actively shape solutions.

Several existing policies and practices offer practical pathways for effective AI governance across contexts: 1. Risk-based regulatory frameworks The EU AI Act introduces a tiered approach, classifying AI systems by risk (e.g., unacceptable, high, limited), with proportionate obligations. This model helps balance innovation with safeguards and can be adapted by other jurisdictions. 2. Algorithmic accountability mechanisms Canada's Directive on Automated Decision-Making mandates impact assessments, audit requirements, and transparency for AI used in government. It provides a replicable template for responsible public sector AI deployment. 3. Ethical and human rights frameworks The UNESCO Recommendation on the Ethics of AI emphasizes inclusion, human rights, and environmental sustainability, offering a normative foundation that countries can localize into policy and practice. 4. Multi-stakeholder research and governance platforms The Global Partnership on AI supports collaborative research on responsible AI, bringing together governments, academia, and industry to generate applied insights and policy guidance. 5. Digital public infrastructure (DPI) approaches India's DPI ecosystem (e.g., digital identity, data exchange frameworks) demonstrates how public digital platforms can enable scalable and inclusive AI applications, particularly in welfare delivery and financial inclusion. 6. Regulatory sandboxes and innovation testbeds Countries like the UK and Singapore have used sandboxes to allow controlled experimentation with AI systems under regulatory oversight helping policymakers learn and adapt regulations in real time. 7. Open-source and responsible AI tools Open models, audit toolkits, and bias detection frameworks (developed by research communities and companies) enable broader access to AI while embedding safeguards. Together, these examples highlight that effective AI governance requires a combination of regulation, institutional mechanisms, technical tools, and collaborative platforms not a single approach.