Global AI Governance Explained: UN AI Panel, Trusted AI Commons & the Global South

Meaning of AI: Artificial Intelligence refers to machine-based systems that perform tasks normally requiring human intelligence — learning from data, recognising patterns, reasoning, making predictions, and generating content such as text, images or code. Modern AI is driven by machine learning and deep learning, in which systems "learn" statistical patterns from very large datasets rather than following hand-written rules.

Why governance now: AI has crossed from narrow, single-task tools into general-purpose, foundation-model systems that touch healthcare, finance, defence, education and public administration at once. Experts compare its transformative potential to that of the steam engine. Because the same model can be used for beneficial and harmful ends, and because a handful of firms and countries control the compute, data and capital required to build frontier models, the technology raises cross-border questions — safety, concentration of power, equity and rights — that no single country can settle alone. This is why 2024–2026 has seen a cluster of new global institutions built specifically to govern AI.

A scientific "reference body", not a regulator: The Panel is the first global scientific body of its kind. Its job is to take a periodic, evidence-based view of the state of the science around AI and to publish an annual report synthesising existing research on AI's opportunities, risks and impacts. Crucially, it is not a regulatory body — it will not set rules, enforce standards or prescribe policy. It deliberately stays out of the politics and policy of AI; that is left to governments in the Global Dialogue.

Composition and independence: It has 40 members drawn from all five UN regions (19 women, 21 men), spanning academia, the private sector, civil society and the technical community. Members serve in their personal capacity (not as government representatives) to protect scientific independence, and the Panel may form working groups and consult external experts.

The IPCC analogy: The Panel is often described as an "IPCC for AI" — just as the Intergovernmental Panel on Climate Change provides governments a common scientific baseline on climate, the IISP-AI is meant to give all countries a shared, impartial evidence base on AI, so that policy debates are grounded in facts rather than hype or fear.

India's seat: Professor B. Ravindran's appointment gives India and the Global South a direct voice in shaping that evidence base, reflecting IIT Madras's work on responsible AI.

The political forum: The Global Dialogue on AI Governance is a universal, inclusive, intergovernmental forum that brings every UN member state to the table, alongside industry, academia, civil society and the technical community. Where the Panel supplies the science, the Dialogue is where governments actually deliberate on governance.

Origin in the Global Digital Compact: Both bodies flow from the Global Digital Compact — the first universal agreement on digital cooperation and AI governance — adopted in September 2024 as part of the "Pact for the Future" at the UN Summit of the Future. The Compact called for inclusive, cooperative governance of AI, and these two mechanisms are how that promise is being operationalised.

First session: The first Global Dialogue is to be held in July 2026 in Geneva, where the Panel will table its inaugural report — the first time the two halves of the architecture (science and politics) formally connect.

This is the core argument made by Prof. Ravindran in his Indian Express interview, viewed from two angles.

The development angle (fragmentation risk): If every country writes its own divergent AI rules, AI development itself becomes fragmented. Companies would have to make their systems comply with many different, sometimes conflicting, requirements across geographies. The likely result is a slowdown in innovation, with developers choosing to roll out services only in "friendly" jurisdictions — leaving others behind.

The data-sovereignty trap: If some countries push data sovereignty to the extreme — insisting that all AI development and infrastructure stay within national borders — it can lead to a dangerous concentration of power, because only a few countries already possess the infrastructure and resources to support an entire AI ecosystem.

The deployment angle (protecting the weak): Many countries in Asia and Africa lack the resources or institutional capacity to frame robust AI regulations that protect their own interests. Without a common floor, they risk being reduced to "digital colonies." A globally agreed set of minimum regulations would protect such countries while keeping innovation alive.

Meaning: "Digital colonialism" describes a situation where developing nations become dependent on foreign-owned AI models, cloud platforms, data pipelines and standards — providing raw data and markets while the value, control and rule-making stay abroad. They consume AI but neither shape it nor capture its gains.

Why developing countries are exposed: Frontier AI requires enormous compute (GPUs), capital, talent and proprietary data — concentrated in a few firms and countries. Nations without these can neither build competing systems nor regulate imported ones effectively, leaving them as price-takers.

Why this matters for India: India frames itself as a "bridge power" between the technology-owning West and the technology-needing South, arguing that AI's benefits must be shared equitably — captured in the summit's invocation of "Sarvajan Hitaya, Sarvajan Sukhaya" (welfare and happiness for all). India offers its Digital Public Infrastructure (DPI) model — Aadhaar, UPI and consent-based data sharing (DEPA) — as a template for affordable, population-scale, public-interest technology.

The legitimate fear: Some argue that because advanced AI could help build the next generation of biological or chemical weapons, the international community should control AI models and tools the way treaties already control biological and chemical weapons.

The worry about "AI non-proliferation": Prof. Ravindran cautions that such controls could mutate into non-proliferation-style language — proposals that only certain "trusted" countries or companies be allowed to develop and use advanced AI freely, while others face restrictions. That would recreate something like the global nuclear order, which many developing countries view as discriminatory (a small club of "haves" and a large group of "have-nots").

The balance to strike: The challenge is to manage genuine catastrophic-misuse risks without entrenching a permanent technological hierarchy — which requires constant vigilance from the Global South in forums like the Dialogue.

A shared toolbox: Announced at the India AI Impact Summit, the Trusted AI Commons is a repository of tools, benchmarks, datasets, evaluation methodologies and protocols for developing and deploying AI safely and responsibly. If, say, a developer wants to test an AI system for agriculture, the Commons would act as a one-stop shop to find what testing tools, benchmarks and protocols already exist.

"Commons" = open and accessible: It is called a "commons" because it is meant to be open, accessible and offered under very liberal licensing, so resource-poor countries are not denied the means to build and test trustworthy AI. To begin with, it is to be hosted and managed by India, drawing on tools already built by bodies such as CeRAI at IIT Madras and by companies like Google. It is one practical building block of the wider global governance structure, launched with 22 partner countries.

The event: Held at Bharat Mandapam, New Delhi (16–21 February 2026), it was the fourth in the global AI summit series and the first hosted by a Global South nation. The series runs Bletchley Park, UK (2023, "AI Safety") → Seoul (2024) → Paris (2025, "AI Action") → New Delhi (2026, "AI Impact") → Geneva, Switzerland (2027).

The shift from "safety" to "impact": Earlier summits framed AI mainly around existential and safety risks. India deliberately reframed the agenda around development, diffusion and real-world impact — structured around three "Sutras" (People, Planet, Progress) and seven thematic working groups ("Chakras"). By moving from harms to benefits, India built a broader tent: the New Delhi Declaration on AI Impact secured support from 92 countries and international organisations, including both the United States and China.

The contested politics: The same US that endorsed the voluntary New Delhi Declaration had, days earlier, voted against the UN Scientific Panel on sovereignty grounds — illustrating a wider tension between non-binding cooperation (which Washington tends to favour) and formal global AI governance structures (which it tends to resist).

The first comprehensive law: The EU AI Act (Regulation (EU) 2024/1689) is the world's first comprehensive, horizontal, risk-based law on AI. It entered into force on 1 August 2024, with prohibited-practice bans applying from February 2025 and full application from 2 August 2026.

A risk-based pyramid: It sorts AI systems into four tiers — unacceptable risk (banned, e.g. social scoring, certain biometric uses), high risk (allowed only after strict conformity assessment), limited risk (transparency duties, e.g. labelling chatbots and deepfakes), and minimal risk (largely unregulated). The EU prioritised AI that is safe, transparent, non-discriminatory and environmentally friendly.

The "Brussels Effect": Because the Act applies to any AI system whose output is used in the EU, and carries fines up to €35 million or 7% of global turnover, it influences company behaviour worldwide — the so-called "Brussels Effect." It illustrates the very fragmentation concern raised above: most other parallel efforts are voluntary, with varying legal force.

The national mission: Approved in March 2024 under the Ministry of Electronics and Information Technology (MeitY), the IndiaAI Mission has an outlay of ₹10,371.92 crore over five years, guided by the vision "Making AI in India, and Making AI Work for India." It is implemented by the IndiaAI Independent Business Division under the Digital India Corporation.

Seven pillars: Compute capacity (10,000+ GPUs), Innovation Centre (indigenous large multimodal and foundational models), Datasets Platform (AIKosh), Application Development Initiative, FutureSkills, Startup Financing, and a dedicated Safe & Trusted AI pillar for responsible-AI tools, governance frameworks and self-assessment checklists. By the 2026 summit, India had provisioned 38,000+ GPUs of sovereign compute, with another 20,000 announced — directly relevant to escaping the "digital colony" trap.

Anchor the science–politics link: A credible inaugural Panel report at the July 2026 Geneva Dialogue can set a shared, hype-free evidence base for all states, especially those lacking in-house technical capacity.

Guard against an "AI non-proliferation" order: The Global South, with India as a leading voice, must ensure legitimate misuse-control does not harden into a permanent two-tier system of AI "haves" and "have-nots."

Build the commons, not just the rules: Operationalising the Trusted AI Commons — open tools, benchmarks and datasets under liberal licensing — is the most direct antidote to "digital colonies."

Strengthen domestic capacity: Scaling sovereign compute, indigenous foundation models and the Safe & Trusted AI pillar under the IndiaAI Mission lets India regulate from a position of strength rather than dependence.

Pursue a minimum-floor consensus: A globally agreed set of minimum regulations can protect weaker states without fragmenting innovation or concentrating power in a few hands.