India has built the digital rails for AI-powered healthcare. The hard part starts now

By Chintan Dave

India’s healthcare AI story is entering a more consequential phase. The question is no longer whether artificial intelligence can read a medical image, summarize a clinical history, flag an unusual insurance claim or help a doctor make sense of complex information. We increasingly know that it can.

The harder question is whether India can turn these capabilities into a trusted, governed and scalable part of healthcare delivery for more than a billion people.

India does not need AI primarily to replace doctors. It needs AI to multiply scarce healthcare capacity: helping clinicians make better-informed decisions, taking elements of specialist expertise closer to underserved communities, detecting disease earlier, strengthening public-health surveillance and making large government health programmes more efficient.

Unlike a few years ago, much of the digital infrastructure needed to attempt this at population scale is beginning to exist.

The digital foundation is becoming enormous
The Ayushman Bharat Digital Mission (ABDM) is creating an interoperable foundation across India’s fragmented health system. As of 12 August 2026, official figures reported 96.43 crore ABHA health identities, more than 110 crore ABHA-linked health records, over 5.47 lakh registered health facilities and more than 10.50 lakh registered healthcare professionals.

These figures matter not merely because healthcare is being digitized. They matter because interoperable, consent-based infrastructure can make information usable across institutions—and, with the right safeguards, can give AI systems a more coherent foundation on which to operate.

Healthcare information has traditionally been scattered across hospitals, laboratories, pharmacies, paper files, PDFs and incompatible systems. Even when technically digital, it is often difficult to exchange or analyze. ABDM’s promise is a different architecture: digital identity, standardized information, interoperability and consent, followed by analysis that supports—not supplants—a human decision.
That architecture could change how care is coordinated. It also raises the stakes for privacy, cybersecurity, data quality and accountability.

AI is already supporting clinical care at scale
One of the clearest examples is eSanjeevani, India’s national telemedicine service. Between April 2023 and November 2025, the Government of India says 28.2 crore consultations benefited from standardized, AI-enabled clinical decision support. The system helps structure data and generate clinical alerts for doctors.

The significance is easy to underestimate. The highest-value role for AI in healthcare may not initially be autonomous diagnosis. It may be helping clinicians navigate information: symptoms, history, medicines, laboratory results, imaging, guidelines and risk factors, often under severe time pressure.

AI can organize that complexity and surface relevant considerations. The clinician remains responsible for the decision. This is a practical division of labour: let machines manage information complexity while humans retain clinical judgment.

Finding disease earlier
Tuberculosis offers another example. The Cough Against TB initiative uses AI-based cough analysis as part of screening. Official figures say more than 1.62 lakh people were screened with the tool from March 2023 through November 2025. The government reported an additional TB case yield of about 12–16% compared with conventional screening in the settings where the tool was used.

The claim should be read as a reported programme result, not as a universal performance guarantee. Even so, the public-health logic is compelling: earlier identification can lead to earlier diagnostic testing and treatment, and may reduce transmission.

The second phase of India’s intensified TB campaign, launched in March 2026, focused on nearly 1.58 lakh high-risk villages and urban wards. This points to a broader shift: analytics can help public-health authorities decide where risk is concentrated, whom to screen first and where limited resources may have the greatest effect.

Taking expertise to where specialists are not
Perhaps AI’s most important contribution in India will come from extending expertise rather than replacing experts. MadhuNetrAI uses AI to support diabetic-retinopathy screening. Government figures indicate that, by December 2025, it had been used across 38 facilities in 11 states, assisted screening of more than 14,000 retinal images and benefited about 7,100 patients.

The broader principle matters more than any single application. A specialist cannot be physically present in every village or primary-health facility. But part of a specialist’s analytical capability can travel digitally: a frontline worker captures an image, software flags abnormalities, and higher-risk cases are prioritized for specialist review.

The goal should not be to build an ‘AI doctor’. It should be to give every healthcare worker access to better intelligence while preserving a clear path to human review.

AI can watch what humans cannot
The same logic applies to disease surveillance. No team can continuously monitor every relevant signal across regions, reports and information systems. Government material reports that a media-based disease-surveillance system generated more than 6,000 AI-assisted alerts as of December 2025. [4]

The useful operating model is simple: signals, machine detection, an alert, human verification and then intervention. An algorithm that produces a flood of irrelevant alerts can make a system worse, not better. AI creates value only when it is connected to a credible human workflow, clear thresholds and feedback on whether alerts were useful.

AI can also protect public money
Healthcare AI is not limited to diagnosis and treatment. By 30 June 2026, Ayushman Bharat Pradhan Mantri Jan Arogya Yojana (AB-PMJAY) had authorized 12.69 crore cashless hospital admissions worth ₹1.92 lakh crore through 37,413 public and private hospitals. A later government factsheet reported more than 40,000 claims processed daily across over 1,900 treatment packages.

At that scale, claims administration becomes an information problem. AI can assist with document analysis, abnormal billing patterns, treatment-guideline checks and indicators of manipulated or suspicious claims. Used responsibly, such systems can help direct public money toward legitimate patient care. Used carelessly, they can also deny or delay valid claims. Human appeal, auditability and measurement of false positives therefore matter as much as detection rates.

Healthcare AI’s impact should be considered across three dimensions: clinical intelligence, operational intelligence and public-health intelligence.

India’s health data could become a research asset
AB-PMJAY alone represents information associated with more than 12 crore hospital admissions, while ABDM is connecting records across a much broader ecosystem. But possessing data is not the same as having research-ready data. Health information is frequently locked inside scans, PDFs, free text and incompatible formats.

AI can help structure that material for disease research, population-health analysis, pharmacovigilance, clinical-trial design and health planning. Yet health data cannot become an unrestricted training pool. Consent, privacy, security, purpose limitation and meaningful governance must remain fundamental.

Bring the algorithm to governed data
In February 2026, the Government of India launched the Benchmarking Open Data Platform for Health AI (BODH), developed by IIT Kanpur with the National Health Authority. The platform is intended to evaluate AI models using diverse, anonymized real-world health data and assess performance, robustness, bias and generalizability before population-scale deployment.

The idea addresses a genuine dilemma. Developers and evaluators need representative data to know whether systems work, but governments and hospitals cannot simply distribute sensitive patient datasets. A governed benchmarking environment offers another model: bring the algorithm to controlled data rather than sending sensitive data to the developer.

If implemented with strong privacy controls, independent evaluation and transparent methods, this could become an important model beyond India.

The biggest challenge is trust
India launched BODH alongside SAHI—the Strategy for Artificial Intelligence in Healthcare for India. SAHI is described by the Health Ministry as a national framework for safe, ethical, evidence-based and inclusive AI adoption, covering governance, data stewardship, validation, deployment and monitoring.
This gets to the central issue. The next challenge is not simply building more models.

It is proving that they deserve to be used. Institutions should be able to answer: Does the system work on Indian populations? Was it tested on representative data? How does it perform across relevant clinical and demographic groups? What are its known limitations? Who remains accountable? Can clinicians override it? Does performance deteriorate after deployment? What happens when the model changes?
These are not abstract ethical questions. They are operating requirements for safe care.

Regulation must deal with a moving target
Conventional medical-device regulation is easier when a product remains substantially unchanged after approval. AI models may change through software updates, retraining, new datasets or altered algorithms. That raises a difficult question: how should a health system authorize something that can continue to evolve?

India’s Medical Devices Rules, 2017 require, where applicable, documentation including software validation, risk-management data and clinical or performance evidence. [7] But healthcare systems will increasingly need explicit controls for significant model modifications after deployment.
The future cannot be ‘approve once and forget’. It must look more like: validate, deploy, monitor, detect change and reassess when necessary.

India needs a healthcare AI assurance layer
Every significant clinical AI system should carry a traceable assurance record: intended purpose; model and version; validation populations; clinical-performance evidence; subgroup performance; known limitations; human-oversight requirements; privacy and cybersecurity controls; approved change boundaries; post-deployment performance; adverse incidents; and evidence of drift.

A credible lifecycle would register the system, classify its risk, validate and benchmark it, clinically evaluate it, deploy it under defined conditions, monitor it and revalidate it when evidence or the model changes.

Technology companies cannot build that ecosystem alone. It requires regulators, clinicians, hospitals, researchers, public-health institutions, standards bodies and independent evaluators.

India has something few countries possess
India’s healthcare challenge is enormous, but the same scale creates an unusual opportunity: nearly a billion digital health identities, more than a billion linked records, hundreds of thousands of registered facilities and professionals, one of the world’s largest publicly financed health-assurance programmes, national telemedicine infrastructure and a growing AI ecosystem.

That combination could make India more than a consumer of healthcare AI developed elsewhere. It could become one of the places where the operating model for responsible, population-scale healthcare AI is developed.

Doing so requires resisting two extremes. The first is fear: refusing useful technology because AI can make mistakes. The second is hype: deploying AI because it is fashionable. Healthcare demands a higher standard. A wrong recommendation can affect a life. That should make institutions disciplined, not passive.

The goal is not artificial doctors
The most useful question for India’s healthcare future is not whether AI will replace doctors. It is what becomes possible when every doctor, nurse, frontline health worker and public-health administrator has responsible access to machine intelligence.

A doctor spends less time navigating records and more time with a patient. A health worker detects a dangerous condition earlier. A TB patient who might otherwise have been missed gets tested. A retinal image taken far from a specialist receives screening assistance. An outbreak signal reaches officials sooner. A suspicious claim is reviewed before public money is lost. A researcher evaluates an algorithm without receiving unrestricted access to patient records. A regulator has evidence to decide whether an AI system deserves to influence care.

That is a more meaningful vision than replacing doctors with algorithms. India’s opportunity is to combine AI, healthcare professionals, digital public infrastructure, Indian data, evidence, governance and trust.

If that combination works, India will not simply have more healthcare AI. It could demonstrate how artificial intelligence can responsibly expand the reach of human expertise across an entire health system.

Verified statistics at a glance
India’s healthcare AI transformation is no longer theoretical; it is operating on population-scale infrastructure. As of 12 August 2026, ABDM had created 96.43 crore ABHA health identities, linked more than 110 crore health records, registered over 5.47 lakh healthcare facilities and more than 10.50 lakh healthcare professionals, while over 1.86 lakh Ayushman Arogya Mandirs were functional nationwide.

AB-PMJAY had authorized 12.69 crore hospital admissions worth ₹1.92 lakh crore as of 30 June 2026, and the government reported processing more than 40,000 claims a day across over 1,900 treatment packages. In AI-enabled delivery, 28.2 crore eSanjeevani consultations benefited from clinical decision support between April 2023 and November 2025; Cough Against TB screened more than 1.62 lakh people and was reported to produce an additional 12–16% TB case yield in deployed settings; MadhuNetrAI assisted screening of more than 14,000 retinal images across 38 facilities in 11 states, benefiting about 7,100 patients; and media-based disease surveillance generated more than 6,000 AI-assisted alerts. The 2026 intensified TB campaign also focused on nearly 1.58 lakh high-risk villages and urban wards.

–Chintan Dave works at the intersection of artificial intelligence, technology, institutional systems and large-scale implementation. He writes about India’s AI ecosystem, responsible AI adoption and the translation of emerging technologies into population-scale impact.