By Amogha Dalvi, Co-founder, AI Leverage
OpenAI announced ChatGPT Health, its consumer-first product, on January 7th. It connects to medical records, understands health history, and navigates the American healthcare system for individual patients. Within a week, Anthropic introduced Claude for Healthcare at the J.P. Morgan conference. An enterprise-first product that is HIPAA-ready, plugged into Medicare databases and fluent in insurance billing codes.
Neither product connects to Ayushman Bharat. Neither reads Hindi prescriptions or Tamil discharge summaries. Neither understands why a patient in Varanasi sees an AYUSH practitioner before consulting with a cardiologist.
OpenAI partnered with b.well to aggregate data from 2.2 million U.S. providers, and Anthropic wired Claude into CMS, a system that doesn’t exist outside America. This is intentional. And it comes with a price tag of ₹399 per month.
OpenAI launched ChatGPT Go in India at a quarter of its U.S. price, free for twelve months. Price below cost, capture the market, raise prices later. The same playbook that bled Ola, forced Flipkart into Walmart’s arms, and kept Zomato unprofitable for a decade. Healthcare could be next.
But India’s government has leverage it isn’t using.
The Ayushman Bharat Digital Mission (ABDM) built a health data stack that rivals anything in the West. ABHA accounts. Consent frameworks. Interoperability standards. As of March 2025, roughly 760 million Indians have ABHA IDs. That infrastructure could be a weapon, but only if someone loads it.
Three policy shifts could change these odds. But all come with caveats.
Firstly, mandate ABDM integration for AI health tools operating at scale in India. If OpenAI wants Indian users, make them connect to Indian systems. This levels the playing field, but only if ABDM is ready to be mandated against. Right now, the stack is promising but immature. If we mandate too early, the compliance costs will strangle the ecosystem. If we mandate too late, the market will likely be already locked.
Next, fund Indian-language medical NLP. The training data exists, fragmented across hospital records, doctor’s notes, decades of vernacular health communication sitting in institutions. What’s missing is aggregation, standardisation, digitisation. Note that this is a privacy minefield and controlling it comes with political temptation. But not getting it done means losing the market to English-first tools that will never fully work here.
Lastly, prioritise procurement towards products tested on Indian patients, designed for Indian workflows in primary care, while keeping channels open for global innovation in tertiary and specialised care. When government hospitals buy AI diagnostics, it must favour tools built for our specific constraints. The problem is that protected markets breed complacency and unprotected markets breed dependency. The question is which gamble is worse.
None of these are guarantees. But consider what we’re protecting.
Indian healthcare isn’t one system. At a government hospital, a single doctor sees 70 patients before lunch. At a private chain, the same procedure costs ten times more. We have nursing homes run by one physician and three relatives. Unregistered practitioners who shouldn’t exist but treat half the country. A patient might visit all five in a single month for the same complaint.
A middle-aged woman in Maharashtra doesn’t start with a doctor. She starts with her pharmacist. Then a WhatsApp forward from her son. A government clinic if symptoms persist. A private hospital only if the clinic fails. Her records either exist on paper she has probably lost, or in her memory, or nowhere.
ChatGPT Health is trained on American patients who see one primary care physician, get referred to specialists, and bill insurance. That logic doesn’t parse here. We need AI that understands why this woman doesn’t come back for the follow-up. Was it cost? Distrust? Distance? Family decision-making? The answer, invisible to an American dataset, is the problem worth solving.
OpenAI’s free year runs out in November 2026. By then, millions of Indian users will have built habits around ChatGPT. Switching costs will compound. Network effects will lock in. The ₹399 price will most likely climb.
The patient who exists in India is multilingual, multi-system, sceptical of formal healthcare, navigating cost constraints and is making decisions with her family. She needs something built for her reality. India has the infrastructure to enable that. The question is whether anyone will act before the window closes.