Blog

India will evolve norms for regulating artificial intelligence, and the AI regulator could function as regulator Securities and Exchange Board of India (SEBI), Prime Minister’s Economic Advisory Council (PMEAC) member Sanjiv Sanyal said on Thursday.

Stating that the model of self-regulation and bureaucratic regulation is unlikely to work in the artificial intelligence space, Sanyal suggested that India could think of having a regulator, which understands the technology and pays attention to how it is evolving.

“You need to create the equivalent (like SEBI) for the AI system. You need a regulator who understands the technology and, instead of wasting time predicting where it will go, pays attention to how it is evolving,” Sanyal said at the ‘DATE with Tech’ event here.

For regulating the AI, a system has to be put in place, you need manual overrides just like circuit breakers in the financial market, he added.

“You need skin in the game and ex-ante accountability. Just like you have boards in the company, who are accountable, you need to make sure that AI and its behaviour… ex-ante, you create skin in the game. You have to enforce regular audits. Just like you need to explain business models and accounts, you need to explain what AI is doing. So, you need standards of explainability,” he said.

Sanyal said all of these things have parallels in the financial world, and this can be applied to the AI evolution and accordingly regulated.

Asked how soon the regulation could come into force, he said, “It (regulatory norms) will evolve. It will happen reasonably fast”.

“India has developed the capability to sensibly manage this space. Once we begin to think along those lines, we will need…good protocol on developing AI regulation,” he said, adding that “more sensible” people around the world now feel that some regulation for AI is needed.

“The question is how do we do it? There are obvious dangers from AI, different countries from the world have tried to do it in different ways,” he said.

The US, he said, has tried a self-regulation model, while China has tried a state-controlled model. Europe is trying a top-down approach to regulating AI.