Why a robust policy framework is essential for AI in India

Globally, AI is seen as the technology that would drive the next wave of economic growth and productivity improvement. Governments and businesses across the globe are keen on leveraging this technology. A recent EY report highlights that GenAI alone could add between US$359 billion and US$438 billion to the Indian GDP¹ in 2030. This is, of course, subject to India implementing an appropriate policy and regulatory framework, as also the degree of AI adoption by companies at scale.  

Building blocks of AI development

AI development has several building blocks. It is driven by technology giants, utilizing deep financial resources, global data access, robust digital infrastructure, immense computing power, specialized talent, and partnerships with research universities.

The regulatory environment in which these giants operate has facilitated innovation and did not come in the way of developing and deploying AI technologies, even though a debate is currently underway. There are learnings that can be leveraged for policy design in India.

India is well positioned to develop and deploy AI

Despite lacking these advantages, the right policy and regulatory push could help India realize its AI adoption potential.  India is now an established global lighthouse for Digital Public Infrastructure (DPI) development through the success of population scale initiatives like Indian Stack, Aadhaar and UPI. These platforms have been developed through specialized organizations, i.e., the UIDAI and NPCI.

Government should create institutional capacity to oversee AI development

A similar approach to AI development could be followed, and the government could consider setting up a specialized agency to lead and oversee this development. Such an agency would leverage the experience of developing and deploying software products at a population scale (and augmenting technical competencies of the Indian private sector), allow for recruitment of specialized manpower and would have access to governmental financial resources. This agency could develop foundational models and  Large language models (LLMs). They could also work on developing used cases for public services and sectors such as education, healthcare, agriculture, smart cities etc., where the government Large language model is the central player.

In parallel, the government needs to take policy and executive actions to facilitate access to the building blocks of AI to unleash the power of individuals, businesses and universities in developing and deploying AI. 

Facilitate access to data

Countries own their data and should use it wisely to benefit from the valuable insights it can provide. Policy actions are required to ensure access to large data sets to start-ups, private corporations, universities, individuals, and the government. The development of Indigenous Training Data Sets (especially for local Indian languages) will be very important. Policy interventions and investments in the creation of structured and unstructured datasets covering both anonymized personal data and non-personal data that are open to the public could be undertaken. Data access could be facilitated by setting up of platforms/ marketplaces/ data trusts.

Facilitate access to digital infrastructure and computing power

Besides data, access to critical digital infrastructure through faster roll-out of 5G, continuing data center development, access to specialized chips and AI specific compute infrastructure is very important. Development of GenAI has gone hand in hand with the increase in computing power. Today, the capability to design and fabricate specialized chips is limited to a handful of corporations and countries only. However, India is home to a large number of technical work force with experience in chip design.

From a longer term and strategic perspective, it is important that India has continuous access to computing power. The government should consider facilitating the development of a chip design company, including through a public private partnership and maybe take inspiration from the success of  Taiwan Semiconductor Manufacturing Company Limited in Taiwan.

Nurturing talent

AI talent is in high demand with a relatively short supply. Skilled AI workforce would be required both within the government for regulatory functions and deployment and in the private sector for both R&D and deployment.  Policies that cultivate and attract specialized talent would need to be enabled.

Regulations may need to be augmented to facilitate AI development

The outcome of the application of a GenAI system cannot be predicted/ determined in advance. In case there is harm, how will one prove causality, especially if the connection between GenAI design and harm is difficult to establish? The decision-making process followed by AI algorithms is opaque and the human mind finds it challenging to understand the process, followed by an algorithm, to reach a certain decision or an outcome.  Hence, the question arises — in case of harm, who would be liable, the developer, deployer or the user?

One school of thought is that the potential harm could be minimized through government oversight over algorithms, such as through standardized testing and risk assessment frameworks. However, government oversight should not become akin to licensing, and stifle innovation. Further, there could be concerns around sharing of confidential data. Then there are issues around how do users know whether the content that they are seeing is generated by an AI system or not? The dilemma is that excessive regulation should not hamper innovation, while at the same time and lack of regulation should not cause a potential harm to hinder development.

For India, an approach that facilitates innovation while managing potential risks is the need of the hour. A regulatory sandbox approach, like the one RBI has deployed with respect to FinTech, could be adopted. If required, new algorithms can be tested in a controlled environment to evaluate the outcomes and access the risks. This approach would help identify necessary regulatory changes and clarify responsibilities in different use cases, such as AI-based medical diagnosis.

Additionally, exploring watermarking technologies, establishing standards, and continuously enhancing understanding of AI implications are crucial steps for a safe AI adoption framework in India.

Upside from AI is unlimited

While this is still a nascent stage for GenAI uptake in India, there is a tremendous sense of optimism in the air. To realize this potential, several things will need to fall into place from a policy perspective. India’s strategic domestic policy interventions may have to be expedited to build competitiveness.