In Geneva on Monday, delegations from across the world gathered for a UN summit convened around a question that sounds deceptively simple: can artificial intelligence benefit all of humanity — safely, fairly, and without causing catastrophic harm? The summit's organizers framed it as a civilizational moment. It is also, beneath the humanitarian language, a contest over who writes the rules of a technology that will restructure labor markets, defense postures, and information ecosystems for decades. These two things — the idealism and the power politics — are not in tension. They are the same negotiation.
Governance as Geopolitics
The architecture of AI governance is emerging as a new theater of norm-setting competition, as consequential as the trade agreements and financial architecture of the 1990s. The European Union's AI Act has established a precautionary, risk-tiered framework that imposes compliance obligations across the value chain. The United States has moved through a combination of executive orders and voluntary commitments from frontier labs, projecting influence through market dominance rather than formal treaty. China has deployed domestic AI regulations while positioning itself internationally as an advocate for sovereignty-preserving approaches that echo its domestic control architecture.
Between these three poles sits a governance vacuum that the Geneva summit is partly designed to fill. How it is filled will determine whether artificial intelligence becomes another domain where the Global South inherits rules written elsewhere, or whether emerging economies with genuine AI capacity claim a seat at the drafting table.
India's interest in this question runs deeper than diplomatic courtesy. The government's IndiaAI Mission carries an outlay of ₹10,371 crore directed at building foundational model capacity, expanding GPU access, and scaling AI applications across public services. That program cannot be evaluated in isolation from the governance environment that will govern how its outputs are deployed internationally, how Indian AI firms access compute infrastructure controlled by a handful of American hyperscalers, and whether open-weight models — which give developing countries the ability to fine-tune and deploy AI without licensing dependency — survive the regulatory cycle or face regulation in the name of safety.
The Fault Line Inside 'Safety'
The word that appears most frequently in UN AI governance discussions is 'safety.' It is also the most contested term at the table. For frontier model developers and their home regulators, safety primarily means catastrophic risk — large-scale autonomous systems acting without human oversight, AI-enabled weapons of mass destruction, recursive self-improvement beyond human control. These are real concerns, and the Geneva summit's warnings about catastrophic harm point in this direction.
But safety, read through a Global South lens, encompasses different harms: AI systems trained predominantly on Western data that misclassify faces, mistranslate languages, or deliver healthcare guidance calibrated to demographics they have never seen. Safety also means the harm of exclusion — governance regimes that mandate compliance infrastructure so expensive that only well-capitalized Western or Chinese firms can afford it, effectively locking out Indian startups and African public health deployments from global AI markets.
This is the structural fault line that analysts working on India's AI positioning have flagged most insistently. When governance norms codify the development practices of incumbent frontier labs as the global standard, they do not merely regulate the market — they freeze it. The compliance costs become a moat. Open-source and open-weight models, which represent the primary mechanism through which countries without sovereign compute infrastructure can build genuine AI capacity, face the highest regulatory pressure precisely because they are the most democratizing. The Geneva process will test whether that pressure translates into binding multilateral norms or whether the summit produces a framework flexible enough to accommodate the full range of AI development approaches.
India's Platform and Its Leverage
India enters Geneva with more substantive standing than it held at comparable technology governance forums a decade ago. During its G20 presidency in 2023, the New Delhi Leaders' Declaration called for globally interoperable AI governance frameworks — language that embedded India's preference for pluralism over regulatory monoculture into the consensus document of the world's largest economic grouping. The Ministry of Electronics and Information Technology has consistently opposed binding international AI regulation in favor of risk-based, context-sensitive national frameworks, a position that preserves regulatory autonomy while allowing India to engage multilateral processes without surrendering domestic policy space.
The more tangible source of leverage is India's digital public infrastructure. The stack built over the past decade — interoperable payments, population-scale identity verification, open e-commerce protocols — is increasingly cited internationally as a governance model for how technology infrastructure can be built with public interest architecture rather than proprietary lock-in. That track record gives New Delhi credible standing when it argues that AI governance need not centralize control in private hands or in a handful of regulatory capitals.
The question is whether India converts that standing into concrete coalition proposals or allows it to dissipate in defensive abstentions. The Voice of Global South Summit framework, through which India convened 123 countries as recently as August 2024, offers a coordination mechanism. A consolidated position among major emerging economies — one that demands mandatory technology transfer provisions, open-weight model access carve-outs for developing countries, and guaranteed Global South representation on any proposed international AI safety body — would carry weight in Geneva that no single country statement can replicate.
The Colonialism Problem
The sharpest version of India's concern centers on what some analysts call AI colonialism: governance regimes that, whatever their stated universalism, operationally entrench the advantages of the companies and countries that built the first generation of large language models. The parallel to earlier technology governance cycles is instructive. Internet governance debates of the 1990s and 2000s produced frameworks that nominally included the Global South while leaving operational control with institutions headquartered in the United States. Intellectual property regimes in pharmaceuticals imposed compliance costs on generic manufacturers that served developing-country populations. The pattern is not inevitable — the Generic Medicines Access campaign eventually produced treaty carve-outs — but it requires deliberate, coordinated resistance at the norm-setting stage, not at the implementation stage when the architecture has already hardened.
India's data governance approach — consent-based, sector-flexible, oriented toward individual rights without adopting the GDPR's extraterritorial architecture wholesale — is precisely the kind of alternative model that, if offered proactively at Geneva, could shape the grammar of international AI governance rather than merely commenting on the version drafted by Brussels and Washington. The institutional vehicle for doing that exists: the UN Secretary-General's AI Advisory Body has called for governance architecture that is not dominated by a handful of advanced economies. India's MEA has supported that call. The next step is converting support into specific textual proposals with coalition backing.
What Comes After Geneva
No single summit rewrites the world order. The Westphalian settlements themselves took years of exhausted negotiation before their procedural principles became the durable grammar of international relations. What Geneva can do is establish the vocabulary — the definitions of harm, the categories of risk, the institutional forms — that subsequent, more binding instruments will inherit. India's interest is in that vocabulary. A governance framework that defines catastrophic AI risk only through the lens of advanced autonomous systems leaves the equally real harms of exclusion, data colonialism, and regulatory capture off the agenda. A framework that defines safety to include equitable access and development-oriented applications gives the IndiaAI Mission and its successors room to operate.
The Geneva summit is, in this sense, less a conclusion than an opening move in a longer norm-setting contest. India's demographic dividend, its AI talent pool, and its ambition to build sovereign AI capacity all depend on getting the opening moves right — which means arriving at Geneva not as a respondent to someone else's draft, but as an author of the framework that follows.
