How Artificial Intelligence is Transforming Organ Transplantation: Promise, Peril, and the Legal Framework in India

ABSTRACT

Artificial intelligence is rapidly entering the field of organ transplantation, from algorithmic donor-recipient matching to machine learning models that predict graft survival and robotic surgical systems that assist in transplant operations. India, which ranks third globally in total organ transplants Mohfw yet suffers a deceased donation rate below 1 per million population, ETV Bharat faces a unique imperative to harness these technologies. This article examines the current and emerging applications of AI in organ transplantation, the legal architecture governing transplantation in India under the Transplantation of Human Organs and Tissues Act, 1994 (THOTA) and its 2011 Amendment, and the ethical challenges posed by algorithmic decision-making in life-and-death medical contexts. Drawing on the 2021 reform of the United States kidney allocation system as a case study, the article argues that while AI holds transformative potential, India’s regulatory framework – including the Digital Personal Data Protection Act, 2023 – remains inadequately equipped to govern algorithmic organ allocation.

The article concludes with policy recommendations for integrating AI into India’s transplant ecosystem while safeguarding equity, transparency, and human oversight.

INTRODUCTION

Every day in India, approximately twenty people die while waiting for an organ transplant.¹ The country performs nearly 20,000 organ transplants annually – a fourfold increase from fewer than 5,000 in 2013 New Kerala Press Information Bureau – yet this meets barely a fraction of the estimated 500,000 patients who need transplants each year.² India’s deceased organ donation rate lingers below 1 per million population (pmp), Mohfw ETV Bharat compared to 49.38 pmp in Spain La Moncloa and 48.04 pmp in the United States.[ Market ^3] Against this backdrop of scarcity, artificial intelligence has emerged as a potentially transformative force. AI-driven systems are already matching donors to recipients in seconds, predicting which grafts will survive a decade after surgery, and identifying potential organ donors in intensive care units before human clinicians do. Yet these technologies raise urgent questions for law and ethics. Who is accountable when an algorithm decides which patient receives a life-saving organ? Can opaque “black-box” models satisfy the requirements of procedural fairness? And does India’s legal infrastructure – anchored in a 1994 statute drafted long before the AI era – possess the tools to govern these systems?

This article explores the intersection of artificial intelligence, organ transplantation law, and bioethics, with India as the primary jurisdiction and global developments as instructive comparisons.

The organ crisis India cannot afford to ignore

India’s transplant crisis is fundamentally a crisis of supply. According to data placed before Parliament, 82,285 patients were on organ transplant waiting lists Vajiram & Ravi as of December 2025, with kidneys accounting for over 60,000 of those cases.[ Vajiram & Ravi ^4] Between 2020 and 2024, 2,805 patients died while waiting StudyIQ Vajiram & Ravi – a figure widely acknowledged as a significant undercount.³ The country witnesses approximately 160,000 road traffic fatalities annually, yet only 1,000 to 1,200 deceased organ donations occur each year.[ Vajiram & Ravi ^6]

The institutional architecture for transplantation rests on the Transplantation of Human Organs and Tissues Act, 1994 (THOTA), which was India’s first legislation to legally recognize brainstem death and regulate organ removal, storage, and transplantation.[ Lippincott Williams & Wilki… ^7] The 2011 Amendment expanded the Act’s scope to include tissues, broadened the definition of “near relatives” to include grandparents and grandchildren, prsindia introduced paired organ exchange (swap donation), mandated transplant coordinators in registered hospitals, mohfw and established the National Organ and Tissue Transplant Organisation (NOTTO) as the apex coordinating body.⁴ NOTTO operates through a three-tier hierarchy: five Regional Organ and Tissue Transplant Organisations (ROTTOs) and approximately twenty-five State Organ and Tissue Transplant Organisations (SOTTOs), collectively networking with over 857 hospitals.[ Indianjnephrol Mohfw ^9]

Despite this institutional scaffolding, India lacks a uniform national organ allocation algorithm. Allocation criteria vary by state, StudyIQ with some employing MELD-based scoring for livers and others relying on first-come or committee-based systems.⁵ This patchwork stands in stark contrast to the algorithmic allocation systems of the United States (UNOS) and Europe (Eurotransplant), where standardised scoring frameworks govern every deceased-donor organ. PubMed Central

AI is already reshaping how organs are matched and outcomes predicted

The most mature application of AI in transplantation is algorithmic donor-recipient matching. The United Network for Organ Sharing (UNOS) in the United States can now match a donated kidney to an eligible candidate in undertwentyseconds.[ UNOS ^11] In March 2023, UNOS launched a “continuous distribution” framework developed in partnership with MIT, employing machine learning and mathematical optimisation to replace the older tier-based system Healthcare IT News with a continuous composite score that eliminates arbitrary geographic boundaries.[ Healthcare IT News ^12]

Predictive analytics represent another frontier. The iBox system, developed by Professor Alexandre Loupy’s Paris Transplant Group, predicts kidney graft loss up to ten years post-transplant Paristransplantgroup using four clinical parameters. Validated across 7,557patients from ten academic centres in Europe and the United States, Paristransplantgroup BMJ Group iBox achieved concordance indices ranging from 0.758 to 0.921 Kidney Medicine and has received qualification opinions from both the European Medicines Agency and the US Food and Drug Administration.[ BMJ Group ^13] A study in Communications Medicine found that iBox outperformed nine transplant physicians at various career stages in predicting long-term graft failure.[ Nature ^14] Similar machine learning models – including the OPOM system for liver allocation, which PubMed Central simulated a 17.6% reduction in waitlist mortality compared to MELD-based allocation PubMed – are advancing toward clinical deployment.[ Science Direct Medscape ^15]

In surgical assistance, the da Vinci robotic system has been used for kidney transplants since 2010, offering three-dimensional visualisation, tremor elimination, and enhanced suturing precision – particularly beneficial for obese recipients.[ PubMed Central ^16] AI-powered imaging tools are also being developed to assess organ viability: deep learning models applied to hyperspectral imaging can score liver viability non-invasively, while computer vision platforms quantify donor liver steatosis to predict early graft dysfunction.[ PubMed Central ^17]

Perhaps most consequential for India’s low deceased donation rate is AI’s potential in donor detection. The German DETECT system, which screens ICU electronic health records every twelve hours for indicators of potential brainstem death, achieved 100% sensitivity in a prospective study PubMed and identified eventual brain-dead patients a median of seven days before formal determination.[ Springer ^18] In Canada, a neural network model using longitudinal laboratory data achieved an AUROC of 0.966 for identifying potential organ donors Nature – and detected donors that the hospital’s own death audit had missed.[ Nature ^19] Given that India’s deceased donation rate remains catastrophically low, the deployment of such AI-based screening in the country’s ICUs could meaningfully expand the donor pool.

Case study:the UNOS kidney allocation reform and the eGFR race correction

The 2021 reform of the American kidney allocation system offers a compelling case study in how algorithmic design choices carry profound equity implications. On March 15, 2021, UNOS replaced its Donation Service Area (DSA)-based allocation system with a proximity-based model known as “acuity circles.” Kidneys are now offered first to candidates within 250 nautical miles of the donor hospital, with proximity points decreasing with distance.⁶ The reform was driven by evidence that the DSA system created a “geographic lottery” – patients in different service areas faced vastly different wait times for identical medical conditions, violating the OPTN Final Rule’s prohibition on residence-based allocation.⁷

The two-year monitoring report showed a 29% increase in overall transplant rates and equity improvements for ethnic minorities, children, women, and highly sensitised candidates.⁸ However, a study in the American Journal of Transplantation found that transplant likelihood for Black candidates was not significantly altered, and cold ischaemia times increased by 12%, raising concerns about organ quality during longer transport.⁹ Running parallel to this structural reform was the eGFR race correction controversy – a striking example of how embedded algorithmic bias can produce systemic harm. The CKD-EPI equation for estimating kidney function had historically included a “Black race” variable that inflated results by 16 to 21 percent, making Black patients’ kidneys appear healthier and thereby delaying diagnosis, referral, and waitlisting.¹⁰ A Yale study estimated that removing this correction would make an additional 31,000 Black Americans eligible for transplant evaluation.¹¹ In July 2022, the OPTN Board mandated race-neutral eGFR calculations, and a subsequent JAMA Internal Medicine study confirmed that the reform produced 5.3 additional transplants per 1,000 Black candidates without reducing transplant access for non-Black patients.¹²

For India, where caste, class, and geographic location profoundly shape healthcare access, the American experience offers a cautionary lesson: algorithms are not neutral instruments. They encode the values and biases of their designers and data.

Ethical fault lines: bias, opacity, and consent

The ethical challenges of AI in organ allocation can be grouped into three categories. First, algorithmic bias: AI systems trained on historical transplant data risk perpetuating existing inequities in healthcare access. As Salybekov et al. observe, such algorithms “often reflect historical inequities” and can produce “discriminatory practices in organ allocation.”¹³ In the Indian context, where urban patients have disproportionately better access to transplant centres and digital health records, AI models could systematically disadvantage rural and lower-income populations.

Second, the “black box” problem: deep learning models used for outcome prediction and matching are frequently opaque. Rueda et al. argue that “the opaqueness of the algorithmic black box and its absence of explainability threatens core commitments of procedural fairness such as accountability, avoidance of bias, and transparency.”¹⁴ When an algorithm determines who lives and who dies, the inability to explain why a particular decision was made poses fundamental challenges to due process.

Third, informed consent: patients on transplant waiting lists may have no knowledge that AI systems influence their priority ranking. PubMed Central MDPI Cohen’s analysis in the Georgetown Law Journal concludes that current legal doctrine does not require disclosure of AI involvement in medical decisions, but argues that the doctrine must evolve – particularly for opaque AI with outsized decisional roles.[ Georgetown Law ^29]

Why AI cannot replace the human transplant committee

Despite AI’s analytical superiority in processing complex multivariate data, organ allocation decisions are not purely technical exercises. They involve irreducibly moral judgements – weighing a child’s claim against an elderly patient’s, balancing medical urgency against

projected survival, and accounting for the social circumstances that data alone cannot capture. A 2023 survey in BMC Medical Ethics found that while 69.2% of respondents found AI in organ allocation acceptable, they expressed concern about the “dehumanisation of healthcare” and questioned whether algorithms could consider “important nuances.”¹⁵

India’s THOTA mandates Authorization Committees Organ India to review unrelated-donor transplant applications India Code and requires brainstem death certification by a panel Lippincott Williams & Wilkins of four physicians.¹⁶ These human-centred safeguards reflect a legislative recognition that transplant decisions carry moral weight that cannot be delegated entirely to automated systems. The appropriate role for AI is as a decision-support tool – augmenting human judgement rather than supplanting it. The “human-in-the-loop” model, where AI generates recommendations that clinicians review and may override, preserves both the analytical power of algorithms and the moral agency of the physician.

The road ahead: building India’s regulatory infrastructure for AI in transplantation

India’s legal framework is not presently equipped to govern AI in organ allocation. The Digital Personal Data Protection Act, 2023 (DPDPA) – India’s first comprehensive data protection statute – does not classify health data as “sensitive personal data,” unlike the GDPR’s designation of health data as a “special category” requiring explicit consent.[ PubMed Central Wikipedia ^32] More critically, the DPDPA contains no provision on automated decision–making analogous to GDPR Article 22, which grants individuals the right not to be subject to decisions based solely on automated processing.[ PubMed Central ^33] For a patient whose position on a transplant waiting list is determined by an algorithm, this is a significant gap.

The EU AI Act (2024), by contrast, would classify organ allocation AI as “high-risk” under both its medical device pathway Easy Medical Device and its public services pathway, imposing obligations for risk management, data governance, human oversight, and transparency.[ Freyr Solutions +2 ^34] India has no equivalent legislation.

Tamil Nadu’s 2021 pilot of an AI-powered organ allocation application – India’s first such initiative – demonstrated both promise and challenges. The system improved data verification and transparency in organ procurement, but transplant teams reported concerns about unintended procedural delays.[ PubMed Central ^35]

Several reforms merit urgent consideration. First, India should develop a uniform national organ allocation algorithm under NOTTO’s coordination, drawing on the continuous distribution model pioneered by UNOS and the ETKAS framework used by Eurotransplant. Second, the DPDPA should be supplemented with sector-specific health data regulations that recognise the heightened sensitivity of transplant data and introduce automated decision-making safeguards – including the right to human review and algorithmic explanation. Third, NOTTO should establish an AI ethics governance committee empowered to audit allocation algorithms for bias, accuracy, and fairness before deployment. Fourth, India should invest in AI– based donor detection systems for ICUs nationwide, modelled on the DETECT system, to address the critical bottleneck of deceased donor identification. Fifth, allocation algorithms and their criteria should be published transparently, enabling scrutiny by clinicians, patients, and civil society. Finally, India’s Ayushman Bharat Digital Mission infrastructure, with over 799 million digital health IDs, provides a foundation for a unified transplant data ecosystem DD News – but only if interoperability standards and data quality protocols are established.¹⁷

CONCLUSION

Artificial intelligence will not solve India’s organ transplant crisis. No algorithm can manufacture consent from grieving families, overcome cultural resistance to deceased donation, or build transplant infrastructure in underserved states. What AI can do – and is already doing elsewhere – is optimise the allocation of scarce organs, predict outcomes with greater accuracy than human clinicians, and identify potential donors who would otherwise be missed. India’s challenge is not merely technological but regulatory. THOTA, drafted in 1994, Bhattandjoshiassociates predates the AI era entirely. The DPDPA, enacted in 2023, lacks the specificity needed to govern algorithmic decision-making DLA Piper in life-and-death medical contexts. The road ahead requires not a choice between human judgement and artificial intelligence, but a legal and institutional framework that harnesses the strengths of both while guarding against the well-documented risks of algorithmic bias, opacity, and inequity. The stakes – measured in thousands of lives lost annually on waiting lists – demand nothing less.

---

ABOUT AUTHOR

Neelu S Rohira is a second year student of LLB at ISBM University Bangalore. The authors research interests include health law, technology regulation, and bioethics. The author maybe contacted at neelusunildas@gmail.com

---

FOOTNOTES

1. Union Health Ministry data submitted to Parliament (2020-2024); reported in Vajiram & Ravi, “India’s Organ Transplant Crisis: Rising Deaths, Long Waitlists,” 2025. ↩
   1. Press Information Bureau, Government of India, “India Landmark Progress in Organ Transplantation,” February 2026; NOTTO Annual Report 2024-25. ↩
   1. Id.; Of the 2,805 deaths, Delhi accounted for 1,425 (nearly half). ↩
   1. Transplantation of Human Organs (Amendment) Act, 2011; PRS Legislative Research, Bill Summary. ↩
   1. Observer Research Foundation, “Inequities, Data Deficiencies, and Capacity Constraints: The Challenges to Organ and Tissue Donation in India,” Special Report. ↩
   1. OPTN, “New Kidney, Pancreas Allocation Policies in Effect,” March 15, 2021. ↩

• 42 CFR Part 121 (OPTN Final Rule); Washington University in St. Louis, “How Kidney Allocation Changed in 2021.” ↩
  • UNOS, “Two-Year Monitoring Report Continues to Show Improvements in Equity,” 2023. ↩
  • Weissberg J. et al., “The Shift to Acuity Circles: Analyzing Changes in Kidney Transplant Access,” American Journal of Transplantation Vol. 25, Issue 8, 2025. ↩
  • The CKD-EPI equation’s race coefficient inflated eGFR by 16-21% for Black patients; STAT News, “Inside the Bruising Battle to Purge Race from a Kidney Disease Calculator,” September 5, 2024. ↩
  • Tsai J., Yale School of Medicine, published in eClinicalMedicine; Yale News, “Abandoning a Race-Biased Tool for Kidney Diagnosis.” ↩
  • OPTN Board Decision, July 27, 2022; JAMA Internal Medicine study, March 2026; STAT News, “Removing Race from Kidney Function Algorithm Helped More Black Patients Access Transplants,” March 10, 2026. ↩
  • Salybekov A.A. et al., “Ethics and Algorithms to Navigate AI’s Emerging Role in Organ Transplantation,” Journal of Clinical Medicine 2025; 14(8):2775. ↩
  • Rueda J. et al., “Just Accuracy? Procedural Fairness Demands Explainability in AI-Assisted Medical Resource Allocation,” AI & Society, 2024; cited in Prince S. and Savulescu J., “When is Black-Box AI Justifiable?” Big Data & Society, 2025. ↩
  • Drezga-Kleiminger et al., “Should AI Allocate Livers for Transplant? Public Attitudes and Ethical Considerations,” BMC Medical Ethics, 2023. ↩
  • THOTA 1994, Section 3(6) (brainstem death certification); Section 9 (Authorization Committee); 2014 Rules (video-recording of proceedings). ↩
  • PIB, “AI in Healthcare Delivery,” February 2026; Ayushman Bharat Digital Mission: 799 million digital health IDs issued as of August 2025. ↩