How well India adopts AI will shape its future

Unlike most artificial intelligence (AI) regulation frameworks, this is not one that is focused solely on risk. On the contrary, we have proposed a two-pronged approach—one that both enables innovation and mitigates risk. In my view, the weaving together of these two seemingly contradictory sets of objectives is what’s novel about this report.

Also Read: Jobs at stake: Legislation could shield employees facing an AI threat

To enable the use of AI in the financial sector, we centred our recommendations around three distinct pillars: infrastructure, policy and capacity. We proposed the establishment of the sort of data infrastructure that would be necessary for building AI models and, more generally, encouraging the use of AI in the financial sector. 

There is a lot of financial sector data in disconnected silos that, if organized, can be helpful for building models and developing AI applications. To that end, we have recommended an initial investment of ₹5,000 crore, with an additional ₹1,000 crore per annum over the next five years.

We have also suggested setting up an AI innovation sandbox in which RBI-regulated entities, fintech firms and just about anyone interested in applying AI innovations to the financial sector can experiment. Since the general-purpose large language models (LLMs) that are currently available may not serve the specialized needs of our financial sector, we have recommended developing sector-specific models capable of catering to the country’s linguistic diversity. 

In order to reward innovation in all these areas, we are hoping that RBI will institute challenge grants and innovation prizes designed to shine a light on noteworthy innovations. Finally, we believe that AI can and should be integrated into India’s digital public infrastructure so that its reach can be extended even further than is currently possible.

Also Read: Artificial workforce: Microsoft’s layoffs are a canary in the coal mine for white-collar jobs

Under the policy pillar, we believe that the country needs a flexible approach that can adapt to keep pace with the evolution of technology. Regulators should periodically assess existing policies and not be afraid to review and amend them if new technologies so demand. 

We have suggested that RBI constitute a permanent multi-stakeholder committee for this purpose. We were also keen to encourage the use of AI to reach those otherwise excluded or underserved. The report encourages the use of AI to bring first-time borrowers into the country’s formal financial system and recommends appropriate relaxation of existing regulations to enable that.

Recognizing that AI systems are fundamentally probabilistic, we believe that it would not be appropriate to apply the binary lens we currently use to determine liability. While we need to ensure that customers are always protected, if we penalize AI developers for each small misstep, we will stifle innovation. 

To that end, we have recommended a graded approach to supervisory action under which the first instance of a failure will not automatically trigger full scope supervisory action, as long as the regulated entity has taken appropriate steps to mitigate harm and proactively remediates issues that have been identified.

Also Read: Rahul Matthan: Technology and law are joined at the hip as they evolve together

There is a need for capacity building in the sector, and for that we have recommended measures to improve the capacity of regulated entities as well as the regulators and supervisors responsible for the sector. 

We have also suggested putting in place a framework for sharing best practices so that the entire sector can benefit from everyone’s experiences with using AI.

As important as it is to incentivize innovation, we must make sure that the framework is designed to mitigate risk. To that end, we have proposed 13 recommendations across three broad pillars: governance, protection and assurance.

Regulated entities are encouraged to establish board-approved AI policies and to ensure that data is protected across the entire AI lifecycle. They need to ensure that they have in place appropriate governance frameworks for the AI systems they use. Since most regulated entities already have product approval processes, we ask that they be modified to take into account the use of AI in their products and services. Beyond pre-release testing, we also recommend regular red-teaming to ensure that failures can be anticipated in advance.

Also Read: Rahul Matthan: Writers should stop worrying and learn to love GenAI

Customer protection is paramount, and regardless of how AI is used  by users, regulated entities must ensure that they are protected at all times. This includes appropriate grievance redressal measures and awareness campaigns to keep customers well-informed while interacting with AI. To ensure AI availability, regulated entities are encouraged to put in place business continuity plans. We have also suggested the establishment of an incident-reporting framework with a good-faith approach to encourage timely disclosure of notable incidents.

AI can prove transformational for India’s development and we have to do whatever is necessary to allow this to happen. It calls for an entirely different approach to regulation than what we currently have—one that encourages responsible innovation and also mitigates risks.

This, in sum, is the Free-AI approach.

The author is a partner at Trilegal and the author of ‘The Third Way: India’s Revolutionary Approach to Data Governance’. His X handle is @matthan.

_{Images are for reference only.Images and contents gathered automatic from google or 3rd party sources.All rights on the images and contents are with their legal original owners.}