Artificial intelligence is advancing faster than most governments can legislate. Every few months, new models, applications and capabilities emerge, creating a difficult policy question: How do you regulate a technology that keeps changing while ensuring that regulation itself does not become an obstacle to innovation?
The European Union’s recent rethink of its artificial intelligence framework offers an important answer: good regulation is not necessarily rigid regulation. It must be capable of evolving.
The EU’s AI framework was designed to establish safeguards around AI according to the level of risk involved. However, as implementation progressed, concerns emerged around administrative complexity, compliance costs, delays in standards and the practical difficulties faced by businesses, particularly smaller ones. The response was not to abandon regulation, but to simplify and recalibrate it.
That distinction holds an important lesson for India.
From Regulation to Adaptive Regulation
The EU AI Act follows a risk-based approach, distinguishing between different categories of AI systems and imposing stronger obligations where the potential risks are greater. Its implementation began in stages, with several important obligations becoming applicable from August 2026.
But regulation written for a rapidly changing technology can quickly become outdated.
The EU therefore introduced its AI Omnibus, which entered into force on July 27, 2026. It seeks to simplify implementation, reduce unnecessary administrative burdens, provide greater legal clarity and give businesses more time to comply with certain high-risk AI requirements.
For example, some high-risk AI rules have been given extended timelines, while regulatory sandboxes and experimentation opportunities have been strengthened. At the same time, the revised framework retains important safeguards relating to safety and fundamental rights.
This is perhaps the most important feature of the European approach:
Simplification does not have to mean deregulation.
The Indian Question: How Much Regulation Is Enough?
India is at a different stage of its AI journey.
It has a massive digital population, a rapidly expanding technology ecosystem and enormous potential for AI in healthcare, agriculture, education, governance, financial services and public administration.
At the same time, AI introduces serious concerns involving privacy, discrimination, misinformation, deepfakes, automated decision-making and accountability.
India therefore faces a delicate balancing act.
Too little regulation could allow harmful applications to spread faster than institutions can respond.
Too much regulation, especially regulation that is complicated or expensive to comply with, could disproportionately affect startups and smaller companies and slow down experimentation.
The objective should therefore not be to create the largest possible rulebook.
It should be to create the most effective rulebook.
Lesson 1: Regulation Must Be Proportionate to Risk
India should avoid treating every AI application as equally dangerous.
An AI system helping farmers analyse crop conditions is fundamentally different from an AI system influencing employment decisions, credit access, policing or critical infrastructure.
A risk-based framework allows policymakers to concentrate regulatory resources where the consequences of failure are greatest.
This would also prevent a situation where low-risk innovators face the same compliance burden as developers of potentially high-impact systems.
Lesson 2: Regulatory Sandboxes Can Become Innovation Laboratories
One of the EU’s important moves has been to expand opportunities for controlled experimentation through regulatory sandboxes.
India can learn from this approach.
Instead of forcing every new AI application to navigate a rigid regulatory environment from day one, selected innovators could be allowed to test technologies under controlled conditions with regulators observing the process.
Such sandboxes can help answer practical questions:
• Does the regulation actually work?
• What risks appear in real-world deployment?
• Which compliance requirements are unnecessarily burdensome?
• What safeguards are genuinely necessary?
In other words, regulators can learn from innovation instead of regulating innovation from a distance.
Lesson 3: Small Companies Need Proportionate Compliance
AI innovation will not come only from large technology corporations.
Startups, universities, researchers and smaller firms will play a major role in developing India’s AI ecosystem.
If compliance becomes excessively expensive, regulation could unintentionally favour large corporations that can afford dedicated legal, technical and compliance teams.
The EU’s recent simplification measures attempt to reduce this problem by extending some proportionate approaches beyond traditional SMEs to small mid-cap companies.
India should similarly ask a simple question whenever it introduces an AI obligation:
Is this requirement actually improving safety, or merely increasing paperwork?
Lesson 4: Simplification Should Never Become Deregulation
This is where India needs to be particularly careful.
Reducing unnecessary paperwork is desirable.
Weakening protections against discrimination, privacy violations, manipulation or unsafe AI is not.
The real objective should be smart regulation: fewer unnecessary procedural burdens but stronger attention to substantive risks.
For example, instead of requiring endless documentation for every AI application, regulators could focus on measurable outcomes such as:
Safety → Transparency → Accountability → Human oversight → Redress
That would make regulation more meaningful rather than merely more complicated.
Lesson 5: Build Institutions That Can Change Their Minds
Perhaps the deepest lesson from the EU experience is institutional.
A regulation should not become sacred merely because it was enacted.
Technology changes. Risks change. Business models change. Therefore, regulatory frameworks must also change.
The EU’s willingness to modify aspects of its AI framework demonstrates an important principle of regulatory maturity: recognising that a rule is not working perfectly is not regulatory failure; refusing to correct it can be.
India should build this adaptability into its AI governance architecture from the beginning.
Periodic reviews, sunset clauses for certain rules, regulatory sandboxes, stakeholder consultations and evidence-based amendments can make AI governance more responsive.
India Should Not Simply Copy the EU
There is another important lesson here.
India should learn from the EU without attempting to reproduce the European model word-for-word.
The EU’s regulatory priorities emerge from its own institutional structure, economic conditions and legal traditions.
India has a different reality.
It has a vast informal economy, enormous linguistic diversity, a huge population of first-time digital users and a development imperative that makes affordable technology particularly important.
Indian AI governance therefore needs to be Indian in design, even when global in outlook.
The country can borrow principles such as risk-based regulation, transparency, accountability and regulatory sandboxes while adapting them to Indian realities.
The Way Forward
India’s AI policy should ultimately rest on three principles:
Innovate freely where the risks are low.
Regulate proportionately where the risks are significant.
Intervene firmly where AI threatens fundamental rights or public safety.
This is a much more sustainable approach than choosing between two extremes: unrestricted technological development on one side and excessive regulation on the other.
The EU’s reset shows that AI governance is not a one-time exercise. It is an ongoing process of experimentation, evaluation and correction.
For India, the challenge is not simply to decide how much AI should be regulated.
The bigger question is:
Can India build a regulatory system intelligent enough to evolve alongside the intelligence it seeks to regulate?
That may ultimately determine whether AI becomes a force that merely transforms India’s economy or one that transforms it safely, inclusively and sustainably.
Comments
Post a Comment