India must govern AI with confidence. Use guardrails, not handcuffs

The good news is that India is not starting from zero. It knows how to innovate privately while also empowering innovation publicly.

Prime Minister Narendra Modi with world leaders during the plenary session at AI Impact Summit 2026, 19 February | ANI

India made a choice in 1991. It chose reform over retreat.

In 2016, it leapfrogged into building Digital Public Infrastructure (DPI). As a result, it is today the world’s most advanced real-time payments market, at about 50 per cent of the global share. And UPI drives over 70 per cent of digital payments, impacting the lives of a billion people.

In this chronology, 2026 is the third momentous occasion. The choice India faces now is whether to govern AI with strategic confidence or bureaucratic fear. That should be the real question at the India AI Impact Summit. If Indians get this wrong, the price to pay will be generational.

India’s choice on AI

AI is already becoming more consequential than the telecom or banking infrastructure. It pervades everything from writing code to diagnosing diseases to optimising logistics—besides, of course, generating persuasive synthetic media at scale. And here lies the dilemma: Should India think short-term—given the immediate pain of transition—or should it plan long-term for its young and ambitious talent?

AI can act as a talent amplifier. It can multiply the output of engineers, civil servants, doctors, and startup founders. It can compress decades of development into years or even months.

While AI amplifies capability and productivity, just like any other tool, it can be misused. The system that improves public service delivery can also be used for mass surveillance. The models that expand access to knowledge can also destabilise information ecosystems. And there is no denying that automation will reshape employment faster than the country’s institutions are prepared to handle. There is a trade-off, and India has a choice to make.

On one end is blind acceleration. Move fast, trust the markets, and clean things up later. This maximises the velocity of innovation, but risks eroding public trust and confidence. 

On the other hand is the conservative approach, the view that new technology shouldn’t be adopted until everything is clear, and unknown things must be restricted. This approach does reduce immediate and visible risks. But it doesn’t eliminate the risk of the loss of opportunity. It entrenches incumbents and deters startups. And it could lead to a future that may be detrimental in the long run.

India cannot afford either of the two approaches. If it takes the slow route, there is a real risk that it ends up becoming one of the downstream consumers of AI systems that are built elsewhere and governed by foreign standards, very similar to what we see in social media today. And over-regulation will suffocate the founders who could define India’s leadership in AI.

A layered framework for India

Governing AI must be a nuanced discussion. It requires a layered framework instead of treating AI as a monolith. There are three key layers to consider:

1. The first layer concerns infrastructure. Large foundation models and high-end compute clusters have an impact on the system. Here, what’s needed is transparency around the data used for training AI models, testing for safety, and safeguarding against misuse. The oversight should match the scale and influence. 
2. The second layer concerns the application. Here, the risk must be contextual. An AI system diagnosing cancer warrants far more scrutiny than one drafting marketing copy. What is required is a sector-specific framework—similar to how banking and medical regulations differ from each other. The accountability and compliance burden must be proportional.
3. The third layer involves harm and liability. Here, the focus should be on measurable harm, such as bias, privacy violations, fraud, and safety failures. Instead of mandating that every algorithm be pre-approved, the focus should be on penalising harmful outcomes (and not responsible experimentation).

Such a layered approach protects citizens appropriately while creating lanes for innovators and builders to experiment and create things that will shape a better future.

The good news is that India is not starting from zero. It has built things at scale from Aadhaar to UPI, and knows how to innovate privately while also empowering innovation publicly. India has proven that DPI can be both inclusive and catalytic—empowering its diverse population while enabling innovators to compete with the best and the brightest.

The same philosophy must be applied to AI. Public compute infrastructure must be made accessible to startups and researchers. High-quality, open data sets must be made available in agriculture, health, climate, and governance. And clear standards must be set for consent, auditability, and accountability.

As a startup founder and investor, I see this as a once-in-a-lifetime arbitrage opportunity. AI flattens geography, but geopolitics raises new walls. So, nations that combine openness with trusted governance are best positioned to lead the next cycle.

I am hopeful that India will get AI right. And when it does, it will not merely adopt global AI standards; it will shape them. It will become a partner of choice for the rest of the world and act as a credible bridge between the United States, China, and the emerging markets. Strategic autonomy in the current age of AI will not come from isolation. It will come from capability coupled with trust.

Confront risks, govern with confidence

There will be risks that must be confronted honestly.

Employment transitions will accelerate at a pace we have never seen before. Re-skilling cannot just be left to a policy makers or the private sector; it must be a continuous AI-enabled national priority.

Information integrity will be tested in a way that even the most digitally savvy professionals will struggle to grapple with. Deepfakes and synthetic media will challenge electoral processes as well as the very fabric of our society, which is based on trust in visuals. Authentication and provenance systems must evolve quickly.

Institutional capacity must also improve. Regulators must have technical literacy. Civil servants must understand data modelling and the associated risks. Governance can’t be outsourced to consultants, and compliance can’t become an afterthought.

As heads of state, policymakers, and innovators gather at the India AI Impact Summit, five principles must guide the conversation: 

1. Adopt risk-based and proportional regulation.
2. Invest aggressively in public AI infrastructure.
3. Mandate transparency for high-impact systems.
4. Create regulatory sandboxes for experimentation.
5. Build AI literacy across government and society.

It is time to craft an India-specific model that is not just principled but also pragmatic and pro-innovation. India must author the future of AI.

By governing AI with confidence, India will not just improve the lives of its citizens or build innovative companies, it will strengthen democratic resilience and create compounding economic leverage for decades.

Manish Maheshwari is a Mason Fellow at Harvard University, where he writes about AI and startups. His X handle is @manishm. Views are personal.

(Edited by Prasanna Bachchhav)