Artificial intelligence now sits at the centre of a global argument that is at once technological, economic, political, and moral. Few technologies have produced such simultaneous excitement and anxiety. To its champions, AI is a general-purpose capability that can transform productivity, accelerate scientific discovery, improve public service delivery, and open new forms of creativity. To its critics, it threatens privacy, jobs, and social trust, and even has the potential to destroy humanity. To governments, it is both an engine of growth and a strategic dependency. To companies, it is not only an unprecedented competitive lever but also a source of legal and reputational risk. And to citizens, it is an increasingly invisible infrastructure: not just a tool one chooses to use but a system quietly shaping information, decisions, and access.
In recent months, this abstract debate has begun to take concrete form. Governments in the US and Europe have tightened export controls on advanced chips, limiting access to geopolitical rivals. The EU has operationalised its AI Act, setting out risk-based regulation that could influence global compliance standards. Meanwhile, companies such as Microsoft, Google, and OpenAI are embedding AI deeply into office software, search engines, and cloud infrastructure, turning it into everyday utility rather than specialised technology. At the same time, public backlash is visible: lawsuits from authors and artists over training data, labour protests over automation, and growing concern about deepfakes in elections.
The “AI debate” is no longer theoretical, and it has moved well beyond whether AI matters. The real issue is who builds it, who controls it, whose values shape it, who benefits from it, and how to build guard rails and govern it. That is why the discussion around AI is no longer only about algorithms. It is about compute, data, energy, talent, standards, institutions, and sovereignty. In other words, AI is becoming an ecosystem contest.
Why it matters for India
For India, this matters more than it may appear to do at first glance. India has immense digital depth, a vast developer base, world-class IT services, rich public digital infrastructure, and one of the world’s largest pools of data-generating users. Yet in most layers of AI, especially the most valuable ones, India remains more consumer than creator. It uses models, chips, cloud layers, and platforms largely built elsewhere. That creates opportunity because adoption can be rapid. But it also creates vulnerability because dependence on external ecosystems in a technology that can shape everything from health systems to military logistics is a strategic risk.
This dependence is already visible in practice. Indian startups building AI platforms and tools often rely on foreign foundational models through API (Application Programming Interface) access. Hospitals experimenting with AI diagnostics frequently deploy imported software, customised locally but not built domestically. Even large IT services firms depend on global cloud providers for compute access. This dependence means that critical capabilities, from pricing to access to uptime, can be influenced by external factors. And this can have a long-term impact on economic value capture and strategic autonomy.
The global debate around AI is often framed through three concerns: privacy, innovation, and governance. These are not separate silos; they feed into one another. Privacy is not simply a rights issue. It affects trust, and trust affects adoption. Innovation is not just about speed; it depends on whether users and firms are willing to share data, experiment, and integrate AI deeply into workflows.
Governance is not merely about restriction; done well, it creates legitimacy, standards, and guard rails that make broader innovation possible. Weak privacy protections can trigger public backlash, which then leads to heavy-handed regulation. Excessively rigid regulation can suppress innovation and push experimentation into opaque spaces. Poor governance can amplify harms and erode confidence in AI-generated outputs, which in turn weakens creative and commercial use. The three concerns are, therefore, mutually reinforcing. The policy challenge is not to “balance” them mechanically but to design systems in which each strengthens the other.
India’s AI opportunity is obvious in sectors such as healthcare, education, agriculture, manufacturing, financial inclusion, and governance. India can use AI to augment medical resources in primary care, personalise learning at scale, improve crop advisories, optimise supply chains, detect fraud, and make public service delivery more targeted and efficient. In a country where human capacity is unevenly distributed but digital scale is enormous, AI can act as a force multiplier. It can extend expertise into places where expertise is scarce. That alone makes AI a national development tool, not just a commercial one.
A worker wearing a GoPro camera on his head records actions through motion capture while folding towels inside a model bathroom at Objectways, an AI data company in Tamil Nadu’s Karur district, on May 13. Feeding first-person footage, called “egocentric data”, into specialised AI models helps robots copy humans.
| Photo Credit:
R. SATISH BABU/AFP
Early experiments illustrate this potential. AI-assisted screening tools are already being tested in public health settings to identify tuberculosis and diabetic retinopathy from medical images. Several edtech platforms are deploying adaptive learning systems that tailor content to a student’s pace and comprehension level. In agriculture, pilot projects use satellite imagery and AI models to provide hyperlocal crop advisories to farmers via mobile apps. In governance, AI is being used selectively for document processing, fraud detection in welfare schemes, and grievance redressal workflows. These examples remain uneven and fragmented, but they demonstrate the breadth of applicability across sectors critical to India’s development.
The challenges
But the challenges are equally significant. The first is infrastructure. AI is not only software; it rests on compute, cloud capacity, data centres, reliable power supply, connectivity, and semiconductor supply chains. India has strengths in software and digital public infrastructure but still lacks sufficient domestic depth in advanced compute and semiconductor manufacturing. Without sustained investment in these layers, the country risks being trapped in downstream applications while the highest-value layers remain elsewhere.
This constraint is increasingly recognised within government and industry. Efforts under the IndiaAI Mission include proposals for shared compute infrastructure and incentives for building foundational models.
Yet, compared with the scale of investment seen in the US and China, where billions of dollars are being poured into developing AI chips and data centres, India’s capacity remains limited. Even where compute is available, access costs can be prohibitive for startups and research institutions, potentially slowing innovation at the frontier.
The second challenge is policy coherence. India’s regulatory approach has often been pragmatic and adaptive, but AI requires clearer rules on accountability, safety, explainability in high-stakes settings, competition, and the use of public data. The absence of a clear framework may encourage rapid experimentation in the short run, but over time it increases uncertainty for firms and leaves citizens under-protected. That can slow adoption in precisely the sectors where trust matters most, such as health, credit, and public administration.
There is some debate about whether India’s light-touch approach risks regulatory lag. While flexibility has enabled rapid innovation in digital payments and identity systems, AI’s potential for systemic harm, ranging from biased decision-making to misinformation at scale, may require more anticipatory governance. The challenge lies in crafting rules that are neither stifling nor reactive but credible and enforceable.
The third challenge is privacy and data governance. AI systems are only as good as the data they are trained on and the contexts in which they operate. India’s scale gives it an advantage in data-rich environments, but data abundance without governance is not an advantage; it is a liability. Consent, anonymisation, data-sharing protocols, security standards, and recourse mechanisms must improve. India cannot build a trusted AI ecosystem if users feel they are merely the raw material for systems they do not understand.
This concern is not merely theoretical. Civil society groups and digital rights advocates have repeatedly highlighted risks around data misuse, inadequate consent frameworks, and the opacity of algorithmic systems deployed by both private firms and state agencies. The implementation of India’s data protection regime will therefore be a critical determinant of public trust in AI systems.
A layered strategy is needed
India needs a layered strategy. It must invest in sovereign or at least trusted compute capacity, support indigenous model development where strategically necessary, create high-quality domain-specific datasets, and build regulatory frameworks that are sector-sensitive rather than one-size-fits-all. It should also use procurement power. The Indian state is a major buyer, convenor, and standard setter. If government ministries, public hospitals, state education systems, and public-sector enterprises specify interoperability, auditability, multilingual support, and privacy-preserving design in AI procurement, they can shape the market in constructive ways.
Public procurement, in particular, could be the decisive lever. India’s experience with digital public infrastructure shows that when the state defines interoperable standards such as in payments or identity, private innovation can build rapidly on top of them. A similar approach in AI procurement could encourage domestic capability building while maintaining openness to global collaboration.
An already distinct approach
India’s approach is already distinct from that of both the West and China, though not always by deliberate grand design. The US has largely relied on private-sector dynamism, venture capital, research depth, and platform power. Its AI ecosystem is innovation-led, firm-driven, and globally influential, though now increasingly shaped by national security concerns. The EU has taken a more rights-based and regulatory approach, emphasising risk categories, accountability, and citizen protections. China, by contrast, combines state direction, industrial policy, platform scale, and centralised data power, all within a framework where strategic control is explicit.
Each of these models also reveals constraints. The US approach has produced technological leadership but has struggled with managing platform dominance and societal risks. The European model offers strong protections but faces criticism for slowing innovation. China’s model enables scale and coordination but raises concerns around surveillance and centralised control.
India sits somewhere else. It has neither the concentrated frontier-model leadership of the US nor the institutional regulatory architecture of the EU, nor the command-and-scale integration of China. What it does have is a unique combination of democratic politics, public digital infrastructure (Aadhaar; UPI, or Unified Payments Interface; etc.), entrepreneurial software capacity, and enormous real-world complexity.
India’s instinct has often been to build enabling rails rather than closed systems: identity, payments, document exchange, and digital service layers that others can innovate upon. This “digital public infrastructure” approach could become India’s signature contribution to AI as well. But it will only work if the AI layer built on top of it remains trusted, contestable, and strategically resilient.
Sundar Pichai, chief executive officer of Alphabet Inc., during the Google I/O Developers Conference in Mountain View, California, US, on May 19, 2026. Alphabet Inc.’s Google unveiled a new design for its iconic search box along with new artificial intelligence coding tools, the latest steps in the company’s multibillion-dollar campaign to extend its influence in the age of AI.
| Photo Credit:
David Paul Morris/Bloomberg
This brings us to a central question: does India need to “own” the entire digital supply chain when it comes to AI? In a literal sense, probably not. No country, not even a major power, can or should seek complete autarky across chips, model architecture, cloud systems, tooling, applications, energy inputs, and talent flows. Total self-sufficiency is neither realistic nor efficient. But strategic self-reliance in critical layers is essential. India must know which parts of the stack are too important to outsource fully. These include compute access, sensitive public-sector deployments, foundational language capabilities in Indian languages, cybersecurity layers, and domain-specific models for sectors tied to welfare, critical infrastructure, and national security.
So, the right goal is not ownership of everything. It is control over what is strategically indispensable, influence over standards and interfaces, and competitiveness in high-value layers where India has a durable edge. This is where the argument for industrial AI becomes compelling. India is unlikely to dominate frontier foundation models at the scale of Silicon Valley or perhaps China. But that does not mean it cannot lead in applied AI. In fact, the largest near-term value may lie not in the biggest model but in the most contextually embedded one.
Industrial AI, enterprise AI, and sector-specific systems may be India’s strongest lane. Why? Because India already has deep capabilities in enterprise software services, process engineering, implementation at scale, and do main adaptation. Combining large language models (LLMs) with proprietary enterprise data, workflow integration, local language interfaces, and measurable business outcomes is a different game from building the next frontier model. It rewards intimacy with the customer, proprietary data, deployment skill, cost discipline, and trust. These are areas where Indian firms can compete effectively. Private enterprise will be crucial here. In the absence of a mature government framework, Indian firms cannot wait passively for regulation to arrive. They should move towards stronger self-regulation and transparency now.
That means clear disclosure of AI use cases, internal model governance committees, documentation of training and fine-tuning processes, bias and security testing, audit trails for high-stakes decisions, and escalation mechanisms when AI systems fail. Companies that treat governance as a strategic asset rather than a compliance burden will be better positioned when regulation eventually catches up. More importantly, they will build the trust needed for deeper enterprise adoption.
There is a broader lesson here. AI governance should be proactive, not reactive. That requires policy frameworks that anticipate categories of harm before they become systemic. India does not need a single monolithic AI law immediately, but it does need a coherent architecture: baseline principles across the economy, with stricter rules for high-risk sectors such as health, finance, employment, and public administration. Sandboxes can help.
Fragmented regulation
Global standards for AI governance are desirable, but full harmonisation is unlikely. Regulation will probably remain fragmented because states differ in political values, industrial interests, legal traditions, and strategic priorities. Even so, fragmentation need not mean chaos. Shared minimum standards on safety testing, model evaluation, data handling, cybersecurity, and synthetic content provenance are highly feasible.
India can play an important role here precisely because it is not identical to any of the major blocs. It can act as a bridge among the innovation-first, regulation-first, and control-first camps. If India can demonstrate how democratic scale, open digital rails, and practical governance can coexist, it could shape global conversations in a way disproportionate to its current position in frontier AI.
Over the next three to five years, Indian AI will likely not win by producing the world’s most advanced general-purpose model. That race is capital-intensive, compute-intensive, and already led by a handful of firms and states. But India can absolutely win in the next layer: multilingual AI, enterprise AI, industrial AI, public-service AI, and domain-adapted models for real-world deployment. It can win where implementation matters more than spectacle.
The most promising path is a stack built around trusted data, sector expertise, and workflow integration. Indian companies can build AI copilots for manufacturing, banking, insurance, logistics, retail, legal processing, coding services, and customer support. Hospitals can deploy clinical-assistance tools adapted to Indian caseloads and languages. Education platforms can create multilingual tutoring systems aligned to curricula and local contexts. Governments can use AI to improve grievance redressal, document processing, fraud detection, and service targeting, provided there is human oversight and due process.
Enterprise AI, especially systems combining LLMs with a company’s proprietary data, is indeed one of India’s strongest opportunities. These systems solve practical problems, create defensible value, and rely less on owning the largest base model. They depend instead on integration, retrieval, workflow design, security, and domain customisation. This suits Indian capabilities. The real moat may not be the model itself but the trusted combination of proprietary data, process knowledge, and deployment competence.
Still, India should resist the temptation to become merely an application layer on top of foreign intelligence infrastructure. If it does that, value capture will remain limited and strategic dependence will deepen. The aim should be a middle path: open to global collaboration but determined to build domestic capacity in critical layers; enthusiastic about innovation but serious about governance; committed to democratisation but aware that democratised access without institutional safeguards can also democratise harm.
The urgency of this choice is becoming clearer with each passing year. As AI becomes embedded in everything from financial systems to public administration and national security, countries that merely consume technology risk locking themselves into structural dependence. Those that build, even to a limited extent or selectively, retain the ability to shape outcomes. India’s path will not be easy, but it will be consequential not only for its own trajectory but for how democratic societies at scale engage with one of the most transformative technologies of the century.
Not just another digital wave
AI is not just another digital wave. It is an operating system for economic and social organisation. That is why the stakes feel so high and why the global conversation swings between excitement and panic. The rewards are transformative; the risks are systemic. For India, the question is not whether to participate. It is whether to participate on terms that preserve agency, build capability, and create broad-based value. If India gets this right, it may not lead every frontier but it can shape one of the most important AI stories in the world: how a large, democratic, developing country uses intelligent systems not only to grow faster but to govern better, include more people, and retain strategic control over the technologies that increasingly run everything.
Vijay Govindarajan is the Coxe Distinguished Professor at Tuck School of Business at Dartmouth College and an NYT and WSJ Best Selling Author.
Also Read | A marvel that’s also a menace
Also Read | The creator is dead
