In India, over 65 per cent of health care spending is out of pocket, and access to qualified medical advice is often a privilege reserved for those in big cities. Only about a fifth of cancers are caught in early stages—when they are most treatable—because screening programmes and radiologists are scarce. Even among patients whose cancers could benefit from targeted therapies such as trastuzumab for HER2-positive breast cancer or tyrosine kinase inhibitors for EGFR-positive lung cancer, only about a third receive the necessary diagnostic tests. India has barely a fifth of the radiologists and pathologists it needs, and even fewer oncology experts. The contrast with western systems, where early diagnosis and precision care are routine, is stark.
Artificial intelligence is often hailed as the great equaliser for such shortages. But we must be clear about what it can—and cannot—do. AI cannot conduct clinical trials, invent drugs, perform surgeries, or offer the empathy that defines care. What it can do is improve prevention, early detection, diagnosis, treatment planning and hospital operations. It can transcribe notes, route patients efficiently, predict resource needs and flag anomalies in scans for doctors to review. Used wisely, AI can be the overworked doctor’s ally and force multiplier, not his or her replacement.
Yet health care is not a market commodity. Education, policing, defence and health care exist for public welfare, not shareholder value. Left to market forces, profit motives can lead to unnecessary procedures, inflated claims and catastrophic medical debt—as the US experience shows. If guided solely by commercial interests, AI-driven health care risks deepening inequities and creating new silos. The government must therefore play a pivotal role, not just as regulator, but as a mission leader ensuring AI serves national health goals first.
The starting point must be a clear definition of health challenges that demand national-scale action. India has proven its capability with polio, smallpox and Covid-19 eradication efforts. The same mission-driven resolve is needed to combat tuberculosis, sickle-cell anaemia, diabetes, cardiovascular disease and our most prevalent cancers. A “problem-first” mindset ensures that AI is used where it truly helps, not where it is easy or fashionable.
We must also chart our own course rather than mimic the west. Their health priorities are shaped by their environments, lifestyles and demographics. Nor should India chase the prestige of massive, resource-hungry AI models designed for publication glory. Just as ISRO succeeded not by being the biggest but the most resource-efficient, India’s AI for health must focus on frugal innovation—smaller, focused models solving real problems.
To power this, India needs a strong data backbone. US researchers rely on The Cancer Genome Atlas, a vast open repository of cancer images and genomic data. India’s department of biotechnology has begun similar initiatives such as the Cancer Image Biobank at the Tata Memorial Centre, but these must scale beyond major hospitals to reflect India’s diversity. Representative datasets spanning geography, hospital type and patient backgrounds will allow AI to learn patterns unique to Indian populations. With scale and diversity, our data can also serve global science, just as India already supplies medical professionals, treatment services and IT expertise to the world.
Responsible data sharing requires balance. Privacy is essential, but excessive fear of data exchange can paralyse research. India’s success with Aadhaar and UPI shows it can build secure, interoperable digital systems at massive scale. Health care deserves similar clarity: anonymised, standardised datasets accessible under strict oversight. The Indian Biological Data Portal and initiatives like the Biotech PRIDE Guidelines and FeED protocol are promising steps in that direction.
Focus, too, is vital. Medical AI cannot thrive if resources are spread thin across general AI research. India needs centres of excellence in medical AI embedded within hospitals, led jointly by clinicians and technologists. Senior doctors must be freed—at least part-time—from routine clinical duties to lead AI initiatives. These centres must also attract skilled AI engineers with competitive pay, autonomy and growth paths. Tata Memorial Centre’s department of digital and computational oncology exemplifies this vision—training oncologists to use and guide AI responsibly. Partnerships like the Koita Centre for Digital Oncology and Koita Centre for Digital Health at IIT Bombay show how public-private collaboration can amplify government initiatives and ensure last-mile reach.
At the heart of this transformation lies trust. Doctors must remain in the pilot seat. AI systems should be explainable, transparent about uncertainty and cautious when data quality is poor, just as a junior doctor knows when to defer to a senior. The goal is not to make doctors complacent or over-reliant, but to strengthen their judgment with reliable, well-calibrated insights.
Trust grows when AI developers work within hospitals rather than apart from them. At Tata Memorial Centre, for instance, predictive models for cancer mutations and survival rates have emerged from close collaboration between oncologists, pathologists and computer scientists. AI adoption, like medicine itself, is built on confidence and shared responsibility.
Still, enthusiasm must be tempered with humility. AI trained on biased or poor-quality data can amplify inequities. Automated predictions should never replace clinical judgment. The promise of AI is not to make medicine more mechanical but more humane.
India has already shown through UPI that interoperability and open standards can transform public systems. If we apply the same mission-oriented design to health care data, AI infrastructure and talent development, we can not only reform our health care system but set a global example for equitable, data-driven medicine.
AI alone will not heal India’s health care. But guided by public purpose, scientific discipline and compassion, it can help deliver what every Indian deserves: timely, trusted and affordable care.
Amit Sethi is a professor of electrical engineering at IIT Bombay, and Dr Swapnil Rane is a professor of pathology and head of digital and computational oncology at Tata Memorial Centre. They are developing open cancer image datasets and AI models for clinical testing, focusing on systems that remain robust to noisy data and cautious in their recommendations when faced with poor quality data or difficult cases.