What Is the SAHI Framework? SAHI stands for Strategy for AI in Healthcare India. It is India’s first national framework dedicated to how artificial intelligence should be developed, evaluated, and used in both clinical care and public health. Developed by the Ministry of Health and Family Welfare (MoHFW) and announced at the Health of India Summit 2026, SAHI is intended to: Define how AI tools can be safely integrated into clinical workflows (OPD, IPD, diagnostics, telemedicine, triage, etc.). Set standards for data use, including consent, anonymisation, and interoperability of health data used to train and run AI systems. Clarify roles and accountability when AI is used in diagnosis or treatment decisions, including what remains the responsibility of the treating physician. Provide a regulatory and ethical guardrail for developers, hospitals, and states adopting AI solutions in healthcare. In practice, SAHI is the reference document that regulators, hospital administrators, and technology vendors will use to decide: Which AI tools can be deployed in Indian hospitals and clinics. Under what conditions they can be used (e.g., as assistive vs. autonomous systems). What documentation, validation, and monitoring are required before and after deployment. For clinicians, this means SAHI will influence: The AI tools that appear in your EMR, PACS, or teleconsultation platforms. How you are expected to review, override, or document AI-generated suggestions. What counts as standard of care when AI is available but not used, or used incorrectly. Subsequent sections of this guide would typically explain: Governance & Regulation – how SAHI aligns with CDSCO, NDHM/ABDM, and data protection rules. Clinical Validation – evidence thresholds, Indian population data requirements, and post-market surveillance. Risk Categories of AI Systems – how tools are classified (low/medium/high risk) and what that means for usage. Data & Privacy – consent models, de-identification, data localisation, and secondary use of data. Liability & Accountability – how responsibility is shared between doctor, institution, and AI vendor. Implementation in Hospitals – procurement, integration, clinician training, and audit trails. Equity & Access – ensuring AI benefits public hospitals, rural health, and primary care, not just tertiary centres. Understanding SAHI is therefore not optional policy trivia; it is central to how you will be expected to practice medicine in an AI-enabled health system in India. SAHI for Clinicians: Key Takeaways and How DocTrust Fits In 1. AI is Assistive, Not Autonomous SAHI clearly states that AI cannot replace clinical judgment. The treating doctor is always the final decision-maker. You cannot shift medico-legal responsibility to an AI tool. If you accept an AI suggestion, you own that decision. Actionable step: In your notes, explicitly document that you reviewed, accepted, modified, or rejected the AI output (e.g., “AI suggestion reviewed; final diagnosis/treatment decided by clinician.”). 2. Indian Data Validation is Non‑Negotiable Any AI tool used in India must be validated on Indian patient data to be SAHI-aligned. Tools trained only on Western datasets may misclassify Indian phenotypes, disease patterns, and presentations. Actionable step: When evaluating AI tools, ask vendors: Has this model been validated on Indian datasets? What were the performance metrics (sensitivity, specificity, PPV, NPV) on Indian cohorts? Is there published or third‑party validation? 3. Risk-Based Classification: Know What You’re Using SAHI follows a risk-tier approach: Low-risk: admin tools (scheduling, billing, reminders). Minimal regulatory burden. High-risk: diagnostic/prognostic tools (cancer detection, sepsis prediction, drug dosing, critical triage). Require validation, audit trails, and post‑market surveillance. Actionable step: If you sit on a hospital committee or advise on procurement, insist that high-risk AI comes with: SAHI-aligned validation documents Clear audit logs Defined post‑deployment monitoring and incident reporting. 4. Evolving Standard of Care: Using AI May Become Expected As validated AI becomes available (e.g., for DR screening, TB on CXR, ECG interpretation), not using it when it’s accessible may be questioned in hindsight. If an AI tool in your hospital flags a finding you miss manually, the issue becomes: should you have used the tool or heeded its alert? Actionable step: Work with your hospital to create clear protocols: When must AI be used? How should discrepancies between AI and clinician be handled? How is this documented in the record? 5. Data Governance & Consent: Align with DPDPA SAHI requires that AI use of patient data respects India’s DPDPA: Inform patients when AI is used in their care. Obtain explicit consent for secondary use (e.g., model training, research). Ensure proper anonymisation/pseudonymisation before data leaves clinical systems. Actionable step: Review your hospital’s consent forms and privacy notices: Do they mention AI use in diagnosis/treatment? Do they cover data use for improving AI models? Are opt‑out mechanisms clear? Specialty-Specific Implications Radiology AI-assisted image reading is already widespread. SAHI will formalize: Use of validated tools with Indian data. Audit logs for each AI inference. Mandatory radiologist sign-off on AI findings. What to do now: Push for platforms that show explainable outputs (heatmaps, bounding boxes, structured reports). Ensure your reporting workflow captures: “AI reviewed; final report by radiologist.” Pathology Whole slide imaging and AI histopathology are high-risk under SAHI. Cancer diagnosis tools must: Be validated on Indian slides. Provide explainable outputs (highlighted regions, feature scores) that a pathologist can confirm or dispute. What to do now: Demand transparent performance metrics on Indian cohorts. Insist on the ability to override AI and record reasons. General Practice & Internal Medicine NMC has already warned against over-reliance on AI chatbots by MBBS students and junior doctors. SAHI reinforces that generative AI (ChatGPT, Gemini, etc.) is not validated CDS. What to do now: Use generative AI only for education, drafting, or brainstorming, not as a substitute for structured clinical decision support. Never document “as per ChatGPT” or similar in clinical notes. Ensure final decisions are based on guidelines, evidence, and your judgment. Telemedicine Platforms like eSanjeevani already use AI triage; SAHI will shape how this scales. Focus areas: rural equity, safe triage, and clear accountability. What to do now: Treat AI triage as a pre-screen, not a final decision. Document your independent assessment, especially when you disagree with AI triage. NMC’s Stance: Caution, Not Rejection NMC (via Dr. Abhijat Sheth, Jan 2026) emphasizes: AI must not replace doctors. Over-dependence is a professional risk. The NBEMS AI course (Dec 2025) signals that AI literacy is becoming part of mainstream CME. SAHI and NMC together: Encourage adoption with safeguards. Keep accountability firmly with clinicians. Push doctors to understand both technical limits and regulatory duties. Practical FAQs for Everyday Practice 1. Does SAHI apply to private hospitals and clinics? Yes. SAHI is a national framework covering both public and private facilities that use AI in clinical care. Enforcement will likely come via regulators like CDSCO and NHA over time. 2. Is there a hard deadline for SAHI compliance? No statutory deadline yet. SAHI is a strategy document, but it is expected to inform future regulations and procurement standards. Professional bodies will likely translate it into practice guidelines. 3. Can I use an AI app on my phone for clinical decisions? You can use it as a reference or educational aid, but not as a validated CDS tool. Acting solely on outputs from unvalidated consumer apps (ChatGPT, Gemini, etc.) creates medico-legal risk and is not SAHI-aligned. 4. What if my hospital buys an AI tool I think is unsafe or inaccurate? Under SAHI, you have both the right and responsibility to raise concerns: Document issues (misclassifications, unsafe suggestions). Report formally to your department, ethics/IT committee, or safety board. Treat it as a patient safety issue, not just a tech complaint. 5. Will AI diagnosis ever carry the same legal weight as a doctor’s? Under current SAHI framing, no. AI is assistive; physician sign-off is mandatory. Legal accountability remains with the treating doctor. How Doctrust Helps with SAHI-Era Documentation SAHI will increase documentation and compliance requirements in the short term: Every AI-assisted decision needs a review trail. Every patient interaction involving AI needs clear consent and disclosure. Hospitals must show audit-ready records for regulators and internal quality checks. Doctrust is built to support this reality: Structured digital prescriptions that clearly attribute decisions to the clinician. Audit-ready OPD and IPD records that can capture: Whether AI was used. What it suggested. How the clinician responded (accepted/modified/rejected). ABDM-compatible workflows that align with national digital health infrastructure, making it easier to integrate with SAHI-compliant AI tools. Templates and fields that can be adapted to: Record AI involvement. Capture patient consent for AI use and secondary data processing. If you want to future-proof your practice for AI-integrated care and stay aligned with India’s evolving health-tech regulations, explore Doctrust at www.doctrust.in. Used correctly, SAHI-compliant AI plus robust documentation via tools like Doctrust can enhance care quality while keeping you on solid professional and legal ground.