Centre for Social and Economic Progress

Considering enormous diversity across understanding of what AI is and what its governance implies, rather than striving for or even expecting a global consensus at the first GDAIG is misplaced, especially amidst the ensuing geo-political scenario. Instead, it should focus on these three outcomes: A) Prioritise objectives of AI Governance and enunciate foundational principles; B) Constitute a multi-stakeholder Working Group on AI Governance (WGAIG) on the lines of Working Group on Internet Governance (WGIG) that was set up by the UN SG in 2003. Its composition must have adequate and effective representation from the global south; C) WGAIG should be tasked with developing taxonomy for AI Governance as well as recommend models and roadmap for its refinement and implementation based on tangible metrics that duly factor in the extant regional, national and sub-national capacities and capabilities but also their needs, aspirations.

• Interoperability of governance approaches
• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI

While all the 7 thematic areas identified by the General Assembly Resolution 79/325 are important, it would be desirable to focus on the four thematic areas suggested above, considering that the global consensus on AI Governance is unlikely to emerge soon and needs further deliberations. Frameworks and other mechanisms for interoperability of governance approaches would provide a pragmatic even if suboptimal solution until a global consensus emerges. Protection and promotion of human rights is at the core of the UN Charter as also within the UDHR and hence, deserves a high priority. Safe, secure and trustworthy AI would foster innovation as well as confidence amongst all stakeholders to design, develop and deploy AI across the broadest spectrum. However, that must be based on social, economic, cultural, linguistic and technical implications of AI. Admittedly, AI capacity-building would be an ongoing work stream, human oversight has its limitations, and 'openness' is not just about political economy but also business models.

Structural exclusion perpetuates the Global South's postcolonial dependency as "rule-takers" who provide training data, deployment sites, natural minerals and other critical elements for this transformative technology, albeit without adequate safeguards or commensurate social and economic gains. To remedy this "wicked problem," AI governance frameworks must be predicated on the trinity of equity, ethics and ecological sustainability. Accordingly, it would be useful to have explicit focus on 'Equity' and 'Ecology'. Focus on Equity is needed to ensure that the Global South and other minorities and vulnerable stakeholders are duly, genuinely and proactively are part of the process rather than being a mere bystander. Focus on Ecology is a must to ensure that AI is resource-efficient and does not lead to catastrophic impact on account of increasing demand of energy, water and critical minerals while also mitigating the challenges such as E-Waste.

In developing countries such as India that continue to have persistent digital divide and low awareness of ecological footprint of AI and often, there is a tendency to welcome AI investments of all sorts without undertaking a comprehensive cost-benefit analysis. Lack of adequate focus on Equity and Ecology would only exacerbate the prevailing inequities and amplify the ecological challenges, even if the impact may not be visible in the short-term.

Though hundreds of AI governance frameworks already exist, their priorities, principles and processes vary significantly even as common taxonomy is amiss. There are also chances of missing the woods for the trees as the prevailing focus is disproportionately on Generative AI and Agentic AI, and not really on AI in a comprehensive manner. This can lead to avoidable confusion and conflict. The forthcoming AI Dialogue represents an opportunity to identify such challenges and rather than being blindsided, can be used to develop the requisite roadmap for future and strive for interoperability frameworks until the global consensus emerges.

Though the AI Dialogue must build upon the existing and ongoing endeavours, the following are worth highlighting: A) New Delhi Declaration on AI Impact, 2026 B) IEEE's Ethically Aligned Design C) UNESCO's AI for Environment and Ecosystems Toolkit for Policymakers

The format and structure of the AI Dialogue can mimic the Internet Governance Forum. However, rather than taking up time for in-person or on-video delivery of prepared remarks, those can be shared in advance and the Dialogue itself must be more interactive and iterative. At the same time, the organisers may like to ask for, encourage and even facilitate thematic discussions on the sidelines and ahead of the AI Dialogue itself. The outcome from such interactions can feed into the Dialogue.

Global South, women, disabled and minorities are underrepresented in global discussions on AI governance. This is not a new phenomenon, however. For example, to mitigate adverse impacts of climate change, the Global South must bear a disproportionate burden of commitments even as the Global North has all but retracted from its legally binding commitments to provide finance and technology under the Paris Agreement. Likewise, the models for financial services, such as for payments from the Global North, are misfits for the majority of the population, and the adoption of certain technologies may lead to vendor lock-in. Short-Term (Up to 2027) Global South countries should create national strategies for digital inclusion and AI adoption by: o aligning with the SDGs; o using multi-dimensional metrics o redesigning institutional framework and policy architecture o prioritizing resource efficiency as well as open technologies, adoption of DPI, state capacity, and leveraging of public-private partnerships; o highlighting the need for building state capacity for AI governance in the World Bank's World Development Report 2026. Medium-Term (Up to 2030) Global South countries should ensure that: o AI infrastructure rollout is accompanied by initiatives to empower communities for innovative AI usage such as preservation of traditional knowledge and support for local languages; o adoption of interoperable trust architectures is deployed with external oversight; o South-South and triangular cooperation frameworks are used for speed, scale and sustainability; and o cybersecurity strategy and data governance frameworks are updated and adapted for the AI era, while also considering their specific needs, resources and capabilities. Long-Term (Up to 2035) Global South countries should: o seek, secure and sustain proportionate representation in transnational and global fora; o strategically partner with willing countries, albeit without compromising their national sovereignty; o inform, influence and internalize frameworks for AI governance

Short video messages, especially if pre-recorded with transcripts translations across languages can be curated in advance, e.g. in case of UNESCO's AI4IA conference. This would help create a buzz around the Dialogue as well as offer more time and space during the Dialogue for forging consensus.

AI solutions need not be complex, costly or demanding extreme computing power or bandwidth. However, they must be contextually relevant and frugal yet smart, prioritising ecological sustainability. Here are couple of examples: Enabling real-time text, speech, and voice translation, Bhashini uses English as the pivot for translation with just 22 language-pair models instead of 231 pairs of one-on-one translation corpora across the 22 official languages in India. Omni-directional microphones at intersections harness honking tendency to detect traffic slowdowns in India with more than 85 percent accuracy, offering a low-cost alternative to expensive and complex cameras and sensors that fail in unmarked, mixed-traffic lanes.