The United Nations Secretary-General on Wednesday welcomed the release of the first comprehensive global scientific assessment of artificial intelligence, produced by an independent expert panel convened under UN auspices. The report maps opportunities, risks, and governance options across AI's full spectrum — from its capacity to accelerate scientific discovery and expand access to public services, to its potential to concentrate power, undermine accountability, and outpace the regulatory instruments states currently possess. The Secretary-General's phrase was direct: the science is here. The harder question — who writes the governance — remains open.
That question is not abstract. Scientific assessments of this kind feed directly into treaty negotiations, UN General Assembly resolutions, and instruments like the proposed Global Digital Compact. Once a multilateral consensus document absorbs a particular risk taxonomy or liability framework, it becomes extraordinarily difficult to revise. The OECD AI Principles, adopted in 2019 and later endorsed by G20 members, show how quickly a rich-country consensus framework becomes universal language. India signed on to those principles during its G20 Presidency and helped shape their final form — but the underlying architecture, the risk categories, the liability assumptions, the definition of high-stakes domains, had already been drafted in Paris and Brussels before New Delhi entered the room.
The Governance Gap Is the Story
What distinguishes this UN assessment from its predecessors is institutional weight. Previous AI governance efforts — the OECD framework, the EU AI Act, the Bletchley Declaration — were either club exercises by wealthy democracies or voluntary codes with no enforcement architecture. A UN expert panel carries the implicit legitimacy of universality, even when its membership skews toward technically advanced economies. Its recommendations, once absorbed into the Global Digital Compact process, acquire the grammar of international law — not binding immediately, but normatively powerful in the way that the Westphalian settlement was procedurally powerful: the framework, once adopted, sets the terms for every subsequent dispute.
The risk that analysts working on Global South digital policy have flagged is precise: if the panel's assessment reflects primarily OECD-aligned use cases — financial fraud detection, autonomous vehicles, social media moderation — the governance architecture it recommends will be calibrated to those contexts. Countries that deploy AI at population scale for public-service delivery, crop-yield prediction, or disease surveillance will find themselves governed by rules designed for a different problem set. The Westphalian analogy holds: the genius of that settlement was that its provisions were procedural, not substantive, leaving states free to organise their domestic affairs. A UN AI framework built around substantive Western risk categories would invert that logic, reaching inside sovereign digital programmes and prescribing their design.
India's Evidence Problem — and Its Solution
India possesses what no other country in the world can claim in the same combination: a biometric identity system at the scale of over a billion enrollments, a real-time payments network processing hundreds of millions of daily transactions, a vaccination management system that coordinated one of the largest public health mobilisations in history. These are not pilot projects. They are production systems, running at a scale that makes India's AI deployment context categorically different from the contexts that dominate multilateral AI discourse.
The structural problem, as analysts at Carnegie India and the Takshashila Institution have consistently noted, is that India has historically under-invested in translating this domestic implementation record into multilateral norm-setting. The gap is not between India's digital ambition and its digital achievement — on that axis, New Delhi's record is strong. The gap is between what India has built and what India argues for in Geneva, New York, and Vienna. Implementation evidence without negotiating presence is a wasted asset. The UN panel's consultation cycle, now open, is the immediate opportunity to close that gap.
The IndiaAI Mission allocates substantial public funding toward sovereign compute capacity, curated datasets, and indigenous model development, giving MeitY a concrete domestic programme to reference in multilateral negotiations. But a budget line is not a position paper. India needs to arrive at the next consultation round with a written submission that foregrounds Digital Public Infrastructure as a proven model — rights-respecting, financially inclusive, democratically accountable — and argues explicitly that any AI governance architecture which cannot accommodate DPI-scale deployments is structurally incomplete.
The Coalition That Isn't Yet Built
India does not need to make this argument alone. Brazil operates Bolsa Família and a national digital payments infrastructure. South Africa is scaling digital identity across the continent's most industrialised economy. Indonesia manages public AI applications across an archipelago of 270 million people. Nigeria is building fintech infrastructure that reaches populations conventional banking never touched. These are not countries that share India's every strategic interest — but they share a structural position: large, fast-deploying digital economies whose governance needs will be systematically misrepresented if the UN panel's policy options are calibrated only to the risk profiles of small, high-income, heavily regulated markets.
Building a Global South AI governance coalition — not as an anti-Western bloc but as a group of countries with documented implementation experience in high-population, resource-constrained contexts — would give India's multilateral AI position the weight that unilateral advocacy cannot. India anchored the Global Partnership for Financial Inclusion during its G20 year and used that platform to advance Digital Public Infrastructure as a global template. The same institutional model applies here. The difference is urgency: financial inclusion norms accumulated over decades, while AI governance frameworks are consolidating in months.
What the Assessment Could Constrain
The specific domains where a restrictive UN AI framework would bite hardest on India's sovereign choices are worth naming. Biometric identification systems linked to benefit delivery sit near the top of every Western AI risk taxonomy. Agricultural prediction models that use satellite and weather data to guide government crop-insurance programmes involve algorithmic decision-making at scale affecting hundreds of millions of farming households. Defence analytics — target identification, logistics optimisation, signals intelligence processing — represent a category that India, like every major military power, is not prepared to submit to multilateral oversight. If the UN assessment adopts the EU AI Act's categorical prohibitions or the US executive order's compute-threshold restrictions as its baseline, each of these domains faces prospective friction.
None of this requires India to oppose AI governance. The opposite is true: a country with India's democratic credentials and its record of building rights-respecting public digital systems is a far more credible voice for responsible AI governance than the corporate lobby groups that have dominated much of the Geneva process. India's argument is not that AI should go unregulated — it is that regulation designed without reference to India's context will govern India badly.
The UN Secretary-General said the science is here. What follows is politics, and in the politics of multilateral norm-setting, presence is the prerequisite for influence. India's digital scale is a fact. Whether that fact becomes a governing principle in the architecture this assessment shapes depends on what New Delhi does between now and when the consultation window closes. The Westphalian delegates built their novel order out of exhaustion and hard-won experience. India arrives at this table with a different asset: the world's most extensive demonstration that AI-adjacent public digital infrastructure can work at democratic scale. That is not a small thing to bring to a negotiation. The question is whether it gets left at the door.
