The transformative potential of artificial intelligence (AI) is becoming increasingly evident, and its impacts are already being felt, touching everything from the daily lives of individuals to broader industries. Consequently, efforts have begun to create global institutions and frameworks to regulate the development and deployment of AI systems.
Some preliminary institutional structures have already been set up. Last year, the United Nations General Assembly established a Global Dialogue on AI in which every country has been invited to participate. An Independent International Scientific Panel on AI is tasked with making periodic scientific assessments to enrich the Global Dialogue discussions.
Professor B Ravindran, head of the Centre for Responsible AI at IIT Madras, is the lone Indian among the 40 people appointed to the Independent International Scientific Panel on AI for a three-year term. He explains the need for a system of global AI governance, why it matters for developing countries like India and what a commons system for AI tools and resources could look like.
What is the International Scientific Panel on AI supposed to do?
This is the first time a global scientific body on AI has been set up. It is supposed to take a periodic view of the state of science around AI, and inform the Global Dialogue on AI.
Our primary responsibility would be to produce reports on the scientific aspects of AI. This body would not get into the politics or policy side of AI discussions. That is the responsibility of the Global Dialogue, in which governments would participate.
Why is a global governance structure on AI needed? Are we moving in that direction?
Speaking in my personal capacity, I feel that having some kind of international agreement on how AI is to be deployed or what kinds of applications we want to use might be useful. As of now, countries are developing their own frameworks to deal with this technology, but that might not be the most effective way forward.
There are two ways of looking at it. One is from the perspective of the development of AI systems. If the regulation is too fragmented, meaning every country has its own set of differing regulations, then the development of AI would also become fragmented. Companies will have to ensure that their systems can adhere to differing sets of requirements in different geographies.
What might happen, as a result, is a possible slowdown in innovation. AI companies might prefer to roll out their services or support innovation only in countries seen as more friendly in terms of their regulations.
AI regulation
At the same time, if some countries begin invoking issues of data sovereignty (seeking control over data originating within their borders) and insist that all AI development and its infrastructure must happen within national boundaries, it can lead to significant concentration of power. Already, a few countries have the infrastructure and resources to support the entire AI ecosystem.
The other perspective is of deployment. Many countries in Asia or Africa might not have the resources or capacities to frame robust regulations to protect their interests. There would always be a danger of them being reduced to digital colonies.
In this context, having a globally agreed-upon set of minimum regulations would be useful. AI is a very transformative technology, with a transformative capability of the order of the steam engine. It would be awful if, owing to a lack of regulatory capacities, certain developing countries were denied the benefits of AI. It will be a disservice to humanity at large.
This is why international agreements and arrangements are important. We need to ensure that the interests of countries that are not AI developers themselves are also protected, and they continue to benefit from advancements in AI.
Would international regulation also control the risks from AI?
There are definite risks, no doubt. People talk about AI’s capability in building the next generation of biological weapons or the next most powerful chemical weapons. We have international agreements and treaties that control biological or chemical weapons. Now, people might also want to bring about international frameworks that control AI models and the tools that enable these weapons to be built.
While that is logical, the fear and worry are that it might take the shape of non-proliferation language. There might be attempts to propose that only certain countries or companies can develop safe and responsible AI, and therefore, they should be the only ones allowed unrestricted development and use of AI systems. That would lead to something similar to the global nuclear regime.
I am worried about the dialogue on AI moving in that direction. Hopefully, it would not happen, but I think we have to be on constant vigil against it.
You were personally involved in developing the framework for Trusted AI Commons, which was one of the main outcomes of the New Delhi AI Impact Summit in February. What is that, and how would it help?
It can be seen as another part of the global governance structure. The Trusted AI Commons is a repository of various tools needed to develop and deploy AI systems in a safe and responsible manner.
Suppose someone wants to test an AI system meant for agriculture. Trusted AI Commons would be like a one-stop shop to figure out what tools are available to test AI deployment, what the benchmarks, datasets and even protocols are. Several resources will be housed in the Trusted AI Commons.
To begin with, this Commons will be hosted and managed by India. The India AI mission has to identify a suitable vehicle to do that. Trusted AI Commons will not commission building these tools right away, but many organisations are already working on it. This includes the Centre for Responsible AI at IIT Madras. Companies like Google have also created their own tools. It is called Commons because it will be open, commonly accessible, with very liberal licensing requirements.
Visitors at the AI Impact Summit at Bharat Mandapam in New Delhi. (Express photo by Praveen Khanna) 
AI regulation
Share
