India needs its own DeepSeek or risks dangerous dependence on US AI models, Bernstein warns

Research firm Bernstein has issued a sharp warning that India must develop its own large-scale AI model — on the lines of China's DeepSeek — or face a future of structural dependence on American AI platforms, with all the strategic, economic and security vulnerabilities that entails. The report, carried by Moneycontrol, frames the challenge not as a matter of national pride but of hard-nosed strategic necessity.

China's DeepSeek captured global attention earlier this year by releasing a highly capable open-weight AI model that matched or exceeded several leading US models on key benchmarks, at a fraction of the reported training cost. The development demonstrated that frontier AI is no longer the exclusive preserve of a handful of US labs — and raised uncomfortable questions for countries like India that have significant AI ambitions but no comparable homegrown model.

Bernstein's concern is that India's current trajectory — which leans heavily on adopting and building applications on top of US-developed models from OpenAI, Google and Anthropic — leaves the country exposed. If access is restricted, prices change, or the models are trained with values and assumptions that do not reflect Indian priorities, Indian businesses and institutions would have limited recourse.

The warning lands as India's government is actively working on an AI policy framework and has allocated funds under the IndiaAI Mission. But analysts have noted that the mission's compute and model-development components remain modest compared to what China or the US are deploying. Bernstein's report is likely to add pressure on the government and private sector to accelerate — and to think bigger.