AI’s rapid rise forces India to rethink ‘light-touch’ regulations

A six-member technology and policy expert committee (TPEC), notified by the government on 13 April, is preparing guidelines that experts believe could see India shun its current “light touch” approach to regulating AI.

Two senior officials familiar with the matter said the recommendations of TPEC, along with those of a 10-member inter-ministerial AI governance and economic group (Aigeg), are likely to significantly differ from the previous AI governance guidelines published by the principal scientific advisor (PSA)’s committee on 15 November. Aigeg was also notified on 13 April. Mint has seen copies of notifications for both groups.

“It’s fair to say that the current approach being considered for various sectors by the TPEC could be different from the light-touch governance guideline that the PSA’s committee had recommended,” the first of the two officials said on condition of anonymity. “But it’s important to note that India’s regulatory policy stance and approach is significantly changing because AI today is itself evolving rapidly—the government never said that its stance on AI regulation will never change.”

India’s “light-touch” approach to AI has so far meant avoiding sweeping laws and regulations on the development and usage of the technology, instead focussing on what's necessary from a user standpoint. For instance, India has so far refrained from working on one law on AI, and has used the existing IT Act and IT rules to regulate issues such as deepfakes.

The need for tighter oversight was felt following recent developments, including two controversies around explicit AI-generated content on Grok, the AI chatbot embedded in social media platform X, the second official said. The launch of advanced models such as US firm Anthropic’s Claude Mythos earlier this month was also a cause of alarm.

“When you see the capabilities that a model like Mythos brings, it becomes clear that this can put various critical sectors at risk, such as financial services or energy. All options will be kept open, including the eventual possibility of an AI law across sectors,” the second official cited above said.

TPEC and Aigeg have both been set up for the long run, and currently do not have a specific timeline for their first report, or even a pre-determined number of reports or issues they will work on, the two people said.

Aigeg, which is the governing body for TPEC, is chaired by union IT minister Ashwini Vaishnaw, while S. Krishnan, secretary at ministry of electronics and IT (Meity) is the convenor of the group. V. Anantha Nageswaran, India’s chief economic advisor, is also part of this group.

TPEC is chaired by Krishnan, and includes five other members—IIT Madras’ data science and AI head B. Ravindran, IIT Gandhinagar’s director Rajat Moona, and members of industry bodies Nasscom, Data Security Council (DSCI) and Manufacturers Association for Information Technology (Mait).

Queries emailed to the IT ministry did not elicit a response till press time.

Stakeholders of the project said the TPEC will seek industry inputs to frame long-term, dedicated AI policies. “There is now broad alignment that AI cannot be approached only through a sectoral lens,” said Ashish Aggarwal, vice-president of public policy at Nasscom. ”As a horizontal technology, it cuts across industries, and while sector-specific regulations will continue to apply, the focus is now on ensuring a coherent and specialized approach to AI governance as a whole."

“A lot of India's digital systems run on infrastructure that can be potentially made vulnerable by unprecedented cyber attack capabilities that foundational models have demonstrated,” Godse said.

Beyond Anthropic's Mythos, there are multiple AI models, including open-source ones with workflows designed to discover and exploit vulnerabilities, that can create catastrophic and cascading impact on critical sectors, Godse added.

“As the groups take shape, structured industry inputs will be critical in shaping policy that is both enabling and practical for India’s AI growth,” Nasscom’s Aggarwal said.

The European Union is the first in the world to have framed a dedicated AI law, with the EU AI Act passed in March 2024. Other geographies, including the US, have legislated AI on a case-by-case basis—an approach that India has also followed so far.