Why S Jaishankar Told the UN Every Country Must Build Its Own AI Future

Why S Jaishankar Told the UN Every Country Must Build Its Own AI Future

Most global tech policy debates feel like two giant powers arguing over who gets to write the rules while everyone else watches from the sidelines. That's exactly the dynamic India's External Affairs Minister S. Jaishankar pushed back against at the United Nations. His core message was blunt. Every nation has the right to shape its own AI ecosystem rather than just consuming tools built elsewhere.

It sounds obvious until you look at how artificial intelligence is actually being built and regulated right now. A handful of companies in Silicon Valley and Beijing control the bulk of advanced foundation models, massive compute clusters, and high-end semiconductor supply chains. When countries outside that inner circle try to join the conversation, they're often handed pre-packaged software and pre-written safety rules. Jaishankar's intervention at the UN Security Council and General Assembly discussions on AI governance wasn't just diplomatic posturing—it was a warning against digital dependency.

What S Jaishankar Actually Argued at the United Nations

When Jaishankar addressed the UN on artificial intelligence, he framed AI not merely as a commercial product, but as a force that will reshape national security, economic productivity, and social stability. His central thesis rests on technological sovereignty. If a country relies entirely on foreign models trained on foreign data, it surrenders control over its own economic destiny and cultural context.

Here is what makes his stance distinct from standard diplomatic speeches. Instead of calling for a total freeze on AI development or backing a single Western or Chinese regulatory framework, India is advocating for a third path. Jaishankar emphasized democratizing access to AI resources while maintaining guardrails against misuse.

Think about what happens when an AI model trained primarily on English-language internet data is deployed in rural India, Sub-Saharan Africa, or Southeast Asia. The model misses local dialects, misunderstands regional legal frameworks, and hallucinates cultural norms. Jaishankar's point is simple. You can't fix those blind spots with top-down rules written in New York or Brussels. Countries need the sovereign capacity to build, train, and audit their own systems.

Why AI Sovereignty Is Suddenly a Frontline Issue

For years, governments treated software like any other imported good. You bought licenses, connected to cloud servers hosted thousands of miles away, and didn't worry much about where the underlying code lived. AI breaks that model completely.

Unlike traditional software, foundation models absorb vast amounts of domestic data and actively shape public opinion, healthcare diagnostics, agricultural planning, and defense logistics. When I talk to policy advisors and enterprise architects working on national cloud deployments, the biggest concern they raise isn't even espionage. It's asymmetric leverage.

If your banking system, public health triage, or power grid optimization runs on an API controlled by a single foreign corporation, a sudden change in export controls, pricing, or terms of service can paralyze critical infrastructure overnight. We already saw glimpses of this during recent semiconductor export restrictions and cloud compute shortages.

The Compute and Data Bottleneck

Building a sovereign AI ecosystem isn't cheap, and pretending otherwise helps nobody. Right now, the global AI race suffers from extreme concentration across three layers:

  • Compute hardware: Specialized GPUs and AI accelerators are heavily concentrated in a few geographies, with lead times and export licenses dictating who gets access.
  • Training data: High-quality, digitized datasets exist overwhelmingly in English and Mandarin, leaving hundreds of major world languages underrepresented.
  • Talent and capital: Sovereign wealth funds and mega-cap tech firms can spend billions on a single training run, a budget that exceeds the entire science ministry allocation of many developing nations.

Jaishankar's remarks directly target this imbalance. By insisting that every nation has the right to shape its AI environment, India is arguing for international cooperation on compute sharing, open-source model weights, and fair data governance so developing economies aren't permanently locked into the bottom tier of the value chain.

India's Playbook for Sovereign AI

India isn't just making speeches at the UN; it's actively testing a blueprint that other nations are watching closely. Over the past decade, New Delhi built what it calls Digital Public Infrastructure (DPI)—systems like Aadhaar for digital identity and UPI for instant payments. Rather than letting private monopolies own the rails, the government built open protocols and let private companies innovate on top of them.

Now, India is applying that exact philosophy to artificial intelligence through the IndiaAI Mission.

Instead of trying to outspend OpenAI or Google on a single monolithic chatbot, the strategy focuses on practical, population-scale utility. Projects like Bhashini are building open-source voice and language translation models across dozens of Indian languages. The goal is straightforward. A farmer who doesn't read or write English should be able to ask a voice interface about crop insurance or weather forecasts in their native dialect and get an accurate answer immediately.

Open Source vs Closed Ecosystems

A major part of shaping a national AI ecosystem comes down to choosing between closed proprietary APIs and open-weight models. Closed models offer convenience and raw frontier performance, yet they act like black boxes. Governments can't inspect how decisions are made or guarantee where sensitive citizen data flows.

Open-weight models—where developers can download, fine-tune, and host the system on domestic servers—give countries real agency. You can strip out irrelevant biases, train the system on local medical or legal records, and keep every byte of data inside national borders. That's why India and several European and Global South partners have consistently pushed back against international treaty proposals that would effectively outlaw or heavily restrict open-source AI development under the guise of safety.

What Most People Get Wrong About AI Governance

When international bodies debate AI regulation, the conversation usually gets hijacked by two extreme camps. One side screams about existential sci-fi risks and wants a global licensing regime run by a handful of superpowers. The other side demands zero regulation so tech giants can move fast and break things.

Both approaches ignore the actual needs of the Global South.

First, AI safety isn't just about rogue superintelligence. For most countries, the immediate threats are deepfakes disrupting elections, automated cyberattacks on financial networks, and algorithmic bias denying people access to welfare or credit. Jaishankar highlighted weaponization and misuse of digital tools as urgent security threats that require multilateral guardrails right now.

Second, regulation without capacity building is just protectionism. If the UN or a coalition of wealthy nations passes strict compliance rules that only trillion-dollar companies can afford to meet, it kills local startups across Asia, Africa, and Latin America before they even start. True governance has to pair safety standards with genuine technology transfer and compute access.

The Hard Realities Developing Nations Face

Advocating for sovereign AI is one thing. Executing it on the ground is messy, expensive, and full of traps. Countries trying to follow the path Jaishankar outlined run into three recurring mistakes.

💡 You might also like: The Iron Ghost of the Steppe

Buying hardware without a data strategy. Several governments have rushed to announce multi-million-dollar GPU purchases, only to let those chips sit underutilized because their public datasets are locked in unformatted PDFs or siloed across uncooperative ministries. Compute without clean data is just an expensive space heater.

Over-regulating domestic startups. In an effort to look proactive, some regulators copy-paste restrictive compliance laws from Europe without realizing those rules crush small local engineering teams that lack armies of compliance lawyers.

Ignoring energy infrastructure. Training and running AI models requires massive, uninterrupted electricity and cooling capacity. You can't build a sovereign data center hub on an unstable power grid.

How Policymakers and Tech Leaders Should Respond

If you're working in technology policy, enterprise strategy, or AI development, the shift toward sovereign ecosystems changes your operational playbook immediately. The era of assuming one global cloud and one global model will serve every market is fading fast.

Start by auditing where your critical data and inference workloads actually live. Relying on a single cross-border API provider creates unnecessary geopolitical and regulatory exposure. Multi-model architectures—where you route routine tasks to locally hosted, fine-tuned open models and reserve frontier commercial APIs for complex reasoning—give you both cost control and resilience.

Invest early in localized datasets. Whether you're building healthcare tools, financial products, or customer support systems, generic global models will continue to underperform in regional languages and local regulatory environments. Curating high-quality domain data is now a stronger competitive moat than writing clever prompts.

Finally, engage directly with emerging national AI missions and public compute initiatives. Governments across Asia, the Middle East, Europe, and Latin America are actively subsidizing local GPU clusters, open-source benchmarks, and sovereign data exchanges. Building on top of those public rails puts your organization in lockstep with where global AI policy is actually heading—toward a multipolar world where every nation demands a seat at the table.

JT

Joseph Thompson

Joseph Thompson is known for uncovering stories others miss, combining investigative skills with a knack for accessible, compelling writing.