Global Development Network

A good outcome would be that countries and companies actually agree on a few clear, practical rules of the road—things like how to test AI systems for safety, how transparent companies should be, and who is responsible when something goes wrong. Not perfect agreement, but enough to show real alignment. Importantly, something tangible should come out of it—maybe working groups, timelines, or joint projects. Without that, it's just talk. Another sign of success would be that the conversation isn't dominated only by a few countries or big tech firms like OpenAI or Google. If smaller nations, especially from the Global South, feel heard and included, the outcomes will carry more legitimacy.

Transparency, accountability, and human oversight;AI capacity-building;Safe, secure and trustworthy AI;Protection and promotion of human rights;

First, safe, secure and trustworthy AI is the most immediate concern. AI systems are already influencing decisions in healthcare, finance, and public services. Without strong safety standards, testing protocols, and safeguards against misuse, these systems can cause real harm. Building trust requires ensuring that AI behaves reliably under different conditions and cannot be easily exploited. Second, transparency, accountability, and human oversight are essential for responsible deployment. Many AI systems operate as "black boxes," making it difficult to understand how decisions are made. This lack of clarity can undermine trust and make it harder to challenge harmful outcomes. Clear lines of accountability-knowing who is responsible when something goes wrong-and maintaining human involvement in critical decisions are key to preventing overreliance on automated systems. Third, protection and promotion of human rights must remain central. AI has the potential to amplify existing inequalities, enable intrusive surveillance, or reinforce bias in areas like hiring or law enforcement. Embedding human rights principles into AI governance ensures that innovation does not come at the expense of privacy, dignity, and equality. Finally, AI capacity-building is crucial for global equity. Many countries lack the resources, infrastructure, or expertise to fully participate in the AI ecosystem. Without targeted efforts to build capacity, the benefits of AI will remain concentrated in a few regions, widening global disparities. Supporting education, knowledge-sharing, and access to tools can help create a more inclusive and balanced development of AI worldwide.

One major gap is AI concentration and market power. A small number of companies and countries currently dominate advanced AI development. This raises concerns about dependency, unequal influence over standards, and reduced competition. Governance discussions should address how to prevent excessive concentration while still encouraging innovation. Another important issue is compute and resource governance. Access to high-performance computing, data, and energy is becoming a key bottleneck in AI development. Questions around who controls these resources-and how they are allocated-have significant geopolitical and economic implications that cut across multiple themes. Environmental impact is also underrepresented. Training and deploying large AI systems can consume substantial energy and water resources. As AI scales globally, its carbon footprint and sustainability implications should be part of governance frameworks, not treated as an afterthought. A further emerging concern is misinformation and synthetic media. Advances in generative AI make it easier to create highly realistic fake content at scale. This affects elections, public trust, and social stability, and requires coordinated responses that go beyond traditional regulatory boundaries. Finally, adaptability of governance systems is critical. AI evolves much faster than most regulatory processes. Static rules risk becoming outdated quickly, so governance models need to be flexible, iterative, and responsive to new developments.

A major challenge lies in safe, secure, and trustworthy AI. India is rapidly adopting AI in areas like fintech, healthcare, and public services, but regulatory frameworks are still evolving. This creates risks of unreliable systems, data misuse, and uneven quality—especially in high-stakes applications like digital lending or diagnostics. In terms of transparency, accountability, and human oversight, many AI-driven decisions remain opaque. For example, algorithmic decisions in hiring or credit scoring can be difficult to audit, raising concerns about bias and fairness. While initiatives like NITI Aayog have outlined responsible AI principles, implementation across industries is inconsistent. Human rights considerations are also critical. With India's large and diverse population, AI systems can unintentionally reinforce linguistic, cultural, or socio-economic biases. The rollout of frameworks like the Digital Personal Data Protection Act, 2023 is a positive step, but enforcement and awareness remain ongoing challenges. At the same time, AI capacity-building presents a major opportunity. India has a strong digital ecosystem and talent pool, supported by initiatives such as Digital India. Expanding access to AI education, infrastructure, and research funding could position India as a global AI leader, particularly in developing inclusive and cost-effective solutions. There is also an opportunity to lead in AI for social good—for example, improving agricultural productivity, expanding healthcare access, and enhancing public service delivery at scale.

First, it can act as a neutral platform for alignment. Countries often approach AI governance with different priorities—innovation, security, or rights protection. The Dialogue can help identify overlapping interests, such as safety standards or risk management, and build consensus without forcing full uniformity. Second, it can support coordination of standards and best practices. Instead of fragmented national rules, the Dialogue can encourage interoperability—shared approaches to testing, auditing, and certifying AI systems. This reduces regulatory friction and helps companies operate responsibly across borders. Third, it can enable knowledge-sharing and capacity-building, especially for developing countries. By connecting governments, researchers, and industry, the Dialogue can facilitate access to expertise, tools, and training. This helps reduce global inequalities in AI readiness and ensures more countries can participate meaningfully in governance discussions. Fourth, it can serve as a bridge between policy and technical communities. AI governance often suffers from a gap between high-level policy discussions and fast-moving technical realities. The Dialogue can bring these groups together, ensuring that policies are informed by technical understanding and that developers are aware of governance expectations

One important foundation is evidence synthesis efforts, such as the Ethics and Governance of AI: A Synthesis Report. These initiatives consolidate global research, stakeholder perspectives, and policy lessons, helping identify common risks like power concentration and governance asymmetries. The Dialogue can amplify such work by turning insights into shared international priorities and policy guidance, rather than leaving them as standalone analyses. Second, the Dialogue can connect to emerging governance infrastructure efforts. Organizations like the OECD emphasize the need for foundational enablers—data systems, digital infrastructure, talent, and institutional capacity—to support trustworthy AI. Similarly, recent work highlights the importance of operational governance systems (e.g., monitoring, auditing, and lifecycle oversight) rather than static principles. The Dialogue can add value by aligning these efforts globally and encouraging interoperable infrastructure standards

To make this effective, the Dialogue should adopt a multi-layered structure—combining high-level plenary sessions for political alignment with smaller, action-oriented working groups focused on specific themes like safety, transparency, and capacity-building. It should also include regional consultations to reflect diverse perspectives, especially from the Global South, and create channels for ongoing input beyond formal meetings (e.g., open consultations or expert submissions). A clear timeline with measurable outputs—such as policy recommendations, pilot collaborations, or shared standards—would help maintain momentum. Finally, establishing a permanent coordination mechanism or secretariat would ensure continuity, track progress, and adapt discussions as AI evolves, turning the Dialogue from a one-time event into a sustained global process.

Global discussions on AI governance often underrepresent voices from the Global South, including countries like India, as well as small and medium enterprises, informal sector workers, marginalized communities (rural populations, linguistic minorities, persons with disabilities), and grassroots civil society organizations. Technical conversations are also frequently dominated by large corporations and a narrow set of experts, leaving out educators, local policymakers, and affected communities who directly experience AI's impacts. To address this, inclusion must go beyond symbolic participation: resources should be provided for meaningful engagement (funding, translation, and capacity-building), and consultation processes should be decentralized through regional forums and local-language dialogues. Mechanisms like participatory policymaking, community impact assessments, and representation in decision-making bodies can ensure these perspectives shape outcomes. In addition, partnerships with local institutions and universities can help bridge the gap between global frameworks and local realities, making AI governance more equitable and context-sensitive.

To move beyond static panel discussions, the AI Dialogue should adopt more interactive and problem-solving formats that actively involve participants. For example, policy labs and simulation exercises can allow stakeholders to collaboratively respond to real-world AI scenarios—such as managing a system failure or regulating a new model—making discussions more practical and grounded. Multi-stakeholder "co-creation sprints" can bring together governments, industry, academia, and civil society to jointly draft guidelines or pilot initiatives within a set timeframe. Open hearings or citizen assemblies, including voices from countries like India, can ensure that public perspectives are directly integrated into decision-making. Digital platforms can also enable continuous engagement, such as crowdsourced consultations, live polling, and asynchronous expert inputs, making participation more inclusive across geographies. Additionally, demo sessions and technical showcases can help policymakers better understand emerging AI capabilities and risks. Together, these formats create a more dynamic, inclusive, and action-oriented Dialogue that encourages collaboration rather than passive participation.