In AI, India’s choice needn’t be between a US leash or Chinese hook

Dhruv Atreja is head of AI research at Fastino Labs in New York, where he has led post-training collaborations with leading US technology companies, working with open-weight models from American and Chinese developers.

On June 12, two of the world’s most capable artificial intelligence models went dark. An American export directive barred foreigners from using Anthropic’s two flagship AI models, and because the company could not check every user’s nationality, it switched them off. Access returned on July 1 but the warning is clear: India’s AI future can be switched off from Washington.

In September, Beijing came calling. At the Brics summit in New Delhi, Chinese President Xi Jinping promised China would “be a pioneer in establishing a Brics AI open-source community”. It is a seductive offer. Chinese open models are capable, cheap and free to adapt; once downloaded, no one in Washington can take them away.

But this is a choice between a leash and a hook. American access can be yanked away overnight. Chinese models cannot be switched off once downloaded, but they arrive with Beijing’s sensitivities built in. Asked whether Arunachal Pradesh is an Indian state, DeepSeek’s chatbot declined to answer. No country of 1.4 billion people should run its banks, courts and public services on either.

Delhi’s response so far has been to chase size. Start-up Sarvam AI plans to build a trillion-parameter model; so does state-backed BharatGen. The government is reportedly exploring an anchor investment of up to 200 billion rupees (US$2.08 billion) in a new frontier AI fund, on top of the 103.7 billion rupees the “IndiaAI Mission” approved in 2024.