SAHI And BODH: AI Framework For Public Healthcare

Policy Update
Aditi Patra

Background

India is the most populated country in the world as of 2026- with 1.47 billion people. While it can be viewed as a strength, it also proves to be its weakness in terms of its healthcare system. With too many people and few doctors, the country has a doctor-population ratio close to 1:1000, with specialist access being worse in villages than in cities. With the advent of technology and artificial intelligence, experts believe there is scope in closing this gap, with the condition that it is carefully rolled out, guided by clear rules on safety and accountability.

Over the years, India has made an effort to equip its healthcare system with AI. In 2018, NITI Aayog’s National Strategy for Artificial Intelligence identified healthcare as a priority sector under its #AIforAll vision. This has led to the building of the digital plumbing required for AI to actually work- such as the Ayushman Bharat Digital Mission (ABDM), which was launched in 2021. It gave every citizen a unique digital health ID (ABHA) and created a system for sharing health records with consent. As of 2026, it has become one of the largest digital health ID systems in the world, with more than 90 crore ABHA IDs being created.

Apart from this, India’s telemedicine platform, ‘eSanjeevani ’, also garnered over 282 million consultations by November 2025. So India already had the data and the digital rails- it was lacking a clear AI rulebook outlining its function assisting this infrastructure.

This was the status quo until 17th February 2026, when the Union Health Minister Shri Jagat Prakash Nadda launched two initiatives together at the India AI Impact Summit in New Delhi: SAHI (Strategy for Artificial Intelligence in Healthcare for India) and BODH (Benchmarking Open Data Platform for Health AI). The former is a policy framework, whereas the latter is a testing platform that puts it into practice. Usually, the norm seen in most countries is to write an AI strategy first and then build its implementation tools as necessary, yet in this case, both were launched together instead of years apart.

Source: Ministry of Health and Family Welfare. (2026). Transforming Healthcare Delivery Through Artificial Intelligence. Press Information Bureau, Government of India.

Functioning

1. What SAHI does

SAHI is a national guiding document rather than a law- it has 5 core principles outlined:

Governance and evidence: AI tools must be properly tested before clinical use
Safe digital infrastructure: secure, well-governed health data systems
Workforce readiness: training doctors, nurses and health workers to work with AI
Ethical oversight: accountability when AI tools cause harm
Equity-centred deployment: making sure AI benefits reach everyone, not just cities

Across these five pillars, SAHI also lists specific recommendations.

2. The “Pro-innovation with guardrails” principle

SAHI states that AI in healthcare should be adopted responsibly, with evidence, safeguards, and human oversight. India’s model allows lower-risk tools such as hospital administration software to scale quickly- due to its priority status-, while reserving stricter checks for high-risk tools like cancer diagnosis systems. Thus, innovation is encouraged while simultaneously following the rules and guidelines.

3. How it fits with existing laws

SAHI is not a law that replaces India’s existing data laws; rather, it acts as an extension of a guiding framework. Decisions like who can access health data, for what purpose, and who is responsible if something goes wrong are already catered to by existing laws. SAHI works alongside such laws like the Digital Personal Data Protection Act, 2023, the National Health Authority’s Health Data Management Policy, and ICMR’s ethical guidelines for AI in healthcare research.

4. What BODH does

On the other hand, BODH was built by IIT Kanpur and the National Health Authority, and has been recognised as a digital public good under ABDM (Ayushman Bharat Digital Mission). It lets AI developers test their models against real, anonymised health data – without that data ever leaving its original location. This protects patient privacy while still checking whether an AI tool actually works well on Indian patients, not just on data from other countries.

5. The regulatory layer

Back in October 2025, CDSCO, which is India’s drug regulator, released draft rules for “Software as a Medical Device” (SaMD), covering AI-based diagnostic tools, alongside SAHI. Its purpose was to sort the software into 4 risk categories (from class A to D), with cancer-detection AI falling in the highest-risk category. SAHI and CDSCO can work in a complementary format, with SAHI providing policy direction, while CDSCO handles medical-device compliance.

Performance

Both SAHI and BODH are in their early days, as they were launched recently in February 2026. Their performance can only be judged through the wider digital health ecosystem, which they are meant to govern.

1. Scale of the digital health backbone: Before AI-specific governance, India’s underlying digital health infrastructure has scaled fast, with ABHA IDs growing from 14.7 crore in 2021 to over 90 crore in 2026.

2. AI already running at scale

After AI tools were added to the National TB Elimination Programme, adverse outcomes reportedly dropped by 27%. eSanjeevani has processed 282 million consultations, of which around 12 million were AI-assisted. The AI-powered Media Disease Surveillance System has issued over 4,500 outbreak alerts. (Ministry of Health and Family Welfare, 2026)

Source: Ministry of Health and Family Welfare. (2026). Transforming Healthcare Delivery Through Artificial Intelligence. Press Information Bureau, Government of India.

3. On-ground pilots

Three institutions: AIIMS Delhi, PGIMER Chandigarh, and AIIMS Rishikesh – have been named Centres of Excellence for AI in healthcare. In December 2025, MadhuNetrAI was launched and is India’s first AI-based community screening programme for diabetic retinopathy, and has already screened over 7,100 patients across 38 health centres.

Impact

India has had data infrastructure in the form of ABDM and data protection rules such as the DPDP Act, but its biggest structural change has been through SAHI and BODH, as they fill a governance gap that simply didn’t exist before. SAHI is the first attempt to address AI-specific risks like biased algorithms or unclear liability when an AI tool gets a diagnosis wrong.

BODH, on the other hand, tackles both performance problems and privacy risk together by letting developers test models against real Indian health data – without moving sensitive records. This is unlike other AI models, which are trained on Western patient data, and often perform poorly on Indian populations- making BODH the perfect solution to a problem that has held back health AI in India for years.

How AI tools are adopted and scaled in practice is addressed by SAHI, which may shape future government procurement and deployment decisions. Hence, it frames responsible adoption as part of a broader ecosystem approach, linking governance, evidence and infrastructure, with workforce readiness.

SAHI is a policy framework which guides implementation, supports accountability and encourages wider use of trustworthy AI in healthcare across different settings. At its launch, the WHO South East Asia office praised India’s approach and viewed SAHI as a useful reference for regional collaboration on healthcare AI.

Emerging Issues

No money attached: SAHI sets out 32 recommendations, but the launch and framework materials do not clearly identify a budget, funding source, or implementation allocation. That makes financing an important follow-up issue.

The 32 recommendations span governance, workforce training, data infrastructure, and procurement – areas that typically require sustained funding. Commentary on the strategy notes mentions blended finance only in passing, with no cost estimates or named funding source, which has been flagged as a structural gap that could limit implementation.

Risk categories are vague: Risk categories are broad. SAHI distinguishes between higher-risk AI uses, such as diagnosis and treatment support, and lower-risk applications, such as record management, but the boundary is not defined in detail. CDSCO’s draft medical-device software guidance is more specific, though the two frameworks still appear to operate in parallel.

CDSCO’s October 2025 draft guidance sets out four defined risk classes (A-D) tied to a device’s intended clinical use, offering more precision than SAHI’s broader risk narrative. However, the guidance itself notes it clarifies existing device rules rather than introducing AI-specific criteria, so how the two frameworks interact for AI tools remains unclear.

Bias risk in BODH’s data: India’s digital health ecosystem still reflects uneven coverage across groups and regions, so benchmark datasets may need careful design to avoid carrying forward existing gaps. If those patterns appear in BODH, model validation could be less representative than intended.

BODH, developed by IIT Kanpur with the National Health Authority, is meant to validate AI tools against real-world anonymised data to reduce bias. Analysts note its usefulness depends entirely on whether the underlying datasets are representative enough, since data adequate for administrative use may not suffice for validating higher-risk clinical tools.

No timeline or targets: SAHI emphasises direction and governance, but the references do not show clear milestones, deadlines, or measurable performance targets. That can make progress harder to track over time.

Reviews of the strategy note that it contains no phased roadmap, milestones, or key performance indicators, positioning it more as a statement of long-term direction than an implementation plan with checkpoints.

States are not equally ready: Since health is a state subject and digital health capacity varies widely across India, implementation will likely differ by state. The references suggest a national framework, but they do not spell out how less-resourced states will be supported.

SAHI was shaped through regional consultations across India and is framed as a shared framework for central and state governments alike. This suggests some awareness of regional variation, but it does not set out differentiated support for states with weaker digital health infrastructure.

Liability remains unclear: SAHI points to responsible use and oversight, but the references do not lay out a detailed liability framework for developers, hospitals, or users. That leaves important accountability questions to be worked out in practice.

SAHI recommends dedicated AI oversight units and clearer human-AI role definitions within institutions, while data handling is separately governed by the DPDP Act and NHA’s Health Data Management Policy. None of these, however, sets out how liability for an AI-related clinical error would be divided among developers, hospitals, and clinicians.

Way Forward

While both SAHI and BODH have made crucial 1st steps, sustained follow-through would be required to translate policy ideas into real-life outcomes. A clear budget and timeline for SAHI’s recommendations, along with the National Health Mission, could strengthen implementation by aligning SAHI’s risk categories more closely with CDSCO’s SaMD classification, so developers work within a clearer regulatory path.

BODH’s datasets would also benefit from published standards for demographic representation, helping ensure that AI tools are tested across India’s diversity rather than only the most digitally connected populations. A simple post-deployment monitoring system, tracking how AI tools perform after they go live, could further improve accountability and public trust.

So what’s missing? Since it’s a relatively new initiative, data on how many AI Tools have been validated through BODH is missing, or the number of states that have adopted SAHI’s recommendations. This is something to track closely over the next year.

Yet dedicated support would most likely be required by states having weaker digital health systems, along with guidance, if SAHI is expected to work across the country beyond only the most advanced states. By following these steps and working on its implementation, SAHI and BODH can prove to be a useful model for how other developing countries approach AI in public health.

References

Central Drugs Standard Control Organisation. (2025, October 21). Draft guidance document on medical device software [PDF]. Government of India. https://cdsco.gov.in/opencms/resources/UploadCDSCOWeb/2018/UploadPublic_NoticesFiles/Draft%20guidance%20document%20on%20Medical%20Device%20Software%2021%2010%202025.pdf

MediaNama. (2026, February 19). How SAHI and BODH shape AI use in India’s healthcare. https://www.medianama.com/2026/02/223-explained-sahi-bodh-ai-use-india-healthcare/

Ministry of Health and Family Welfare. (2026, February 17). Union Minister launches SAHI and BODH initiatives [Press release]. Press Information Bureau, Government of India. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2229226

Ministry of Health and Family Welfare. (2026). Strategy for Artificial Intelligence in Healthcare for India (SAHI). Ayushman Bharat Digital Mission Portal, Government of India. https://abdm.gov.in/sahi

Ministry of Health and Family Welfare. (2026, February). Transforming healthcare delivery through artificial intelligence [PDF]. Press Information Bureau, Government of India. https://static.pib.gov.in/WriteReadData/specificdocs/documents/2026/feb/doc2026213788701.pdf

National Health Authority. (2026). Ayushman Bharat Digital Mission crosses 90 crore ABHA accounts. Digital Health News. https://www.digitalhealthnews.com/ayushman-bharat-digital-mission-crosses-landmark-milestone-of-90-cr-abha-accounts

World Health Organisation, South-East Asia Regional Office. (2026, February 17). Launch of the Strategy for AI in Healthcare for India (SAHI). https://www.who.int/southeastasia/news/speeches/detail/launch-of-the-strategy-for-ai-in-healthcare-for-india-(sahi)

Asia Actual. (2025, November 24). India releases draft guidance on medical device software. https://asiaactual.com/blog/india-releases-draft-guidance-on-medical-device-software/

ICTworks. (2026, February 24). SAHI: Radical artificial intelligence for health framework from India. https://www.ictworks.org/sahi-radical-artificial-intelligence-for-health-framework-from-india/

MediaNama. (2026, February 19). How SAHI and BODH shape AI use in India’s healthcare. https://www.medianama.com/2026/02/223-explained-sahi-bodh-ai-use-india-healthcare/

India Corporate Law (Cyril Amarchand Mangaldas). (2026, January 29). Medical device as software: Has CDSCO guidance changed the rules? https://corporate.cyrilamarchandblogs.com/2026/01/medical-device-as-software-has-cdsco-guidance-changed-the-rules/

About the Contributor

The author is a Sociology Honours student at Jesus and Mary College, Delhi University, with research and writing experience across social impact, market research, and internet governance (Youth IGF India, APIGA India 2026). Her interest lies in more ethical, unbiased frameworks in tech policy and data governance, as well as in gender studies.