‘Neither feasible nor desirable’: Why AI slowdown debate looks different in India

analysis Asia ‘Neither feasible nor desirable’: Why AI slowdown debate looks different in India Fears that increasingly powerful artificial intelligence could pose an existential threat have fuelled calls for a slowdown. India, however, is not yet pushing the AI frontier, making wider adoption - and managing its consequences - a more immediate concern, say analysts. BENGALURU, India: Farm robots, drones and diagnostic tools filled a sprawling exhibition, alongside displays showing how artificial intelligence could be put to work in classrooms and public services. The showcase at the India AI Impact Summit in New Delhi in February offered a glimpse of the AI future the country is chasing - one focused less on pushing the boundaries than on putting the technology to work across its vast economy. That puts India in a different position as calls grow globally to slow the development of the world’s most powerful AI models, say analysts. They say the world’s sixth-largest economy is not yet pushing that frontier, making wider adoption and its potential economic gains a more immediate priority. The bigger calculation is how to harness AI for productivity and development while managing the disruption it could bring to jobs, observers say. “Slowing adoption is neither feasible nor desirable (for India) … for most of India’s workforce, the risk is too little diffusion, not too much,” Ravi Venkatesan, founder of the Institute for Human Flourishing, an India-based research and policy institute, told CNA, referring to the spread and adoption of AI. At the same time, moving faster will require New Delhi to better track where jobs are being lost or transformed and draw clearer guardrails around high-stakes uses of AI, experts said. A DIFFERENT DEBATE FOR INDIA Calls to slow the development of frontier AI have intensified amid warnings that increasingly capable systems could escape human control and, in the worst case, pose an existential threat to humanity. The push has heavyweight backing from within the industry, with Anthropic’s Dario Amodei, OpenAI’s Sam Altman and SpaceX’s Elon Musk among those calling for the frontier to be paced. Those concerns have been sharpened by recent incidents in which advanced AI agents went beyond their assigned tasks. Among the latest to come to light was an OpenAI experimental model hacking an Australian government health data portal while undergoing internal training. But India has indicated that this is not the race it is currently running. Sources in India’s Ministry of Electronics and Information Technology said there was “no reason” to pause the country’s AI research because it is not at the frontier of model development, local news agency Press Trust of India (PTI) reported on Sep 21. “The issue is coming up with frontier models,” the ministry sources told PTI, adding that the government’s focus is instead on applications and models suited to Indian needs. The global debate over slowing AI development has largely been shaped by countries and companies developing the most powerful models, Nidhi Singh, an associate fellow at Carnegie India, the local arm of the global public policy think-tank, told CNA. For India and much of the Global South, the more pertinent question is how to use AI capabilities across agriculture, education, public services and other parts of an economy where adoption remains uneven, said Singh. Venkatesan from the Institute for Human Flourishing said India currently imports frontier AI capability, relying on advanced models developed by foreign companies rather than producing comparable systems at home. India does have homegrown AI models, including those developed by Sarvam AI and the government-backed BharatGen initiative, but they remain far smaller in scale than the leading global frontier systems. “Domestic restraint would not stop global firms deploying AI to Indian customers. It would only slow down Indian firms and workers from capturing the gains,” he said, adding that informal and small businesses would stand to lose the most. At the same time, analysts pointed out that moving quickly on AI is also a competitive necessity for India’s technology industry. Achyuta Ghosh, executive research leader at HFS Research, a global technology and business advisory firm, said Indian IT services companies and global capability centres (GCCs) - in-house technology and operations hubs set up in India by multinational companies - are under pressure to use more AI because their global clients already expect faster work at lower cost. Clients are also increasingly pushing for contracts tied more closely to outcomes rather than the amount of labour involved, Ghosh told CNA. If Indian providers and GCCs fail to keep pace, some work could shift to foreign competitors that automate faster, be brought back in-house by clients using AI or even absorbed by software platforms, he said. “The practical choice facing India concerns the terms on which it takes part in AI-led services,” Ghosh said. India’s IT and business-process management industry generated about US$315 billion in revenue in FY2026. The country also has 2,117 GCCs employing about 2.36 million people, according to a 2026 report by India’s technology industry association and consulting firm Zinnov. THE ENTRY-LEVEL SQUEEZE While the economic case for India to keep adopting AI is strong at both ends of the economy, the labour-market impact is unlikely to be felt evenly, experts warned. In sectors where AI adoption is moving fastest, the first sign of disruption may be fewer openings for fresh graduates and younger workers rather than large-scale layoffs, Singh from Carnegie India said. Yet the bigger concern is what happens when the entry-level roles that once trained young workers begin to disappear, industry analysts said. India’s technology sector has long relied on a pyramid model in which large numbers of graduates enter through testing, maintenance, support and basic coding before moving into more complex work, said Ghosh. “The pyramid has been both a business model and, in effect, a national skilling system,” he said. That model is already starting to change, according to HFS Research. Ghosh said AI-heavy projects are being run with leaner teams while campus hiring has become smaller and more selective. GCCs, drawn by the country’s large, lower-cost talent pool, are now favouring smaller teams and more experienced workers, he said. Ghosh said the shift also complicates the common argument that AI will simply augment workers rather than replace them. An individual engineer may indeed become more productive with AI, while the company ultimately needs fewer people to get the job done. That also means corporate promises to reskill workers tell only part of the story, he pointed out. “The public language of reskilling describes the intent more accurately than it describes the full reality of deployment,” he said, adding that firms are increasingly using attrition, slower replacement and fewer new positions to move towards leaner teams. Jaspreet Bindra, co-founder of AI&Beyond, an AI education and advisory firm, said the less visible consequence is the weakening of career ladders. Entry-level jobs in coding, customer support, research and back-office work have traditionally allowed young employees to build skills and judgment through experience, he told CNA. If AI absorbs more of those tasks, companies could end up demanding experienced workers without maintaining enough of the junior roles through which that experience was gained, Bindra added. “(India needs) transition support, apprenticeships, portable benefits and incentives for firms to preserve paid pathways into skilled work, not merely training courses,” he said. Ghosh said that some companies are experimenting with alternatives, including apprenticeship-style programmes, university partnerships and junior roles built around refining and supervising AI output, but such efforts remain small compared with the scale of the change. MANAGING THE AI SHIFT If India plans to keep accelerating AI adoption, it will need much better data on who is benefiting and who risks being left behind, analysts said. Singh from Carnegie India said the country’s main labour-force survey is not designed to capture how AI is changing specific occupations, while many widely used measures of AI exposure are based on the US economy. India needs its own methodology to track how jobs and tasks are changing, she said, particularly as disruption may show up through slower hiring and changing roles long before it appears as mass unemployment. Ghosh said the Indian government and industry have put far more energy into “AI infrastructure, compute, and skilling announcements”. “Much less into measuring what is happening to jobs and planning the transition for the people affected,” he said. But better measurement does not mean every use of AI should be allowed to move at the same pace, experts cautioned. “India should not slow general-purpose adoption, but it should slow deployment where error, opacity or concentration can permanently harm people,” AI&Beyond co-founder Bindra said, citing areas such as hiring, credit, welfare, healthcare and workplace surveillance. Venkatesan from the Institute for Human Flourishing said a key goal should be to ensure AI helps small businesses and self-employed workers earn more, rather than simply becoming more efficient. For example, AI tools could help small businesses find more customers, qualify for loans and navigate regulatory requirements, he said. Venkatesan suggested using existing networks such as self-help groups, farmer organisations, business associations, local accountants and government agricultural advisers to help small businesses and informal workers adopt AI, rather than relying on direct-to-user apps. The government could also use AI to support decision-making by frontline health workers and teachers, rather than for surveillance or to replace staff, Venkatesan said. “India cannot shape the technological frontier, but it can and must shape the adoption pathway,” he said.