Clinical AI Diagnostic Systems and Medical Negligence: Allocating Tort and Criminal Liability under Section 106 BNS

I. I. The Algorithmic Transformation of Clinical Medicine in India

The integration of artificial intelligence into clinical medicine represents one of the most promising yet legally fraught frontiers of technological innovation. Deep learning convolutional neural networks (CNNs) and multi-modal foundation models are now routinely deployed across Indian healthcare institutions for automated radiological image interpretation (MRI, CT scans, and X-rays), real-time diabetic retinopathy screening, oncological pathology grading, and robotic surgical assistance. In many controlled benchmarks, clinical AI systems achieve diagnostic accuracy rates comparable or superior to experienced specialists.

However, the translation of statistical predictive models into messy, high-stakes clinical realities introduces unprecedented liability dilemmas. Machine learning algorithms can fail catastrophically when encountering out-of-distribution clinical presentations, image noise, rare demographic pathology variations, or hardware sensor drift. When an AI diagnostic software misses an aggressive malignant lesion or an algorithmic dosing calculator recommends a lethal pediatric narcotic concentration, who bears legal liability?

In Indian jurisprudence, resolving healthcare AI failures requires navigating a complex intersection of criminal negligence under Section 106 of the Bharatiya Nyaya Sanhita 2023, civil medical negligence under the Consumer Protection Act 2019, statutory medical device regulations, and product liability. This treatise analyzes physician liability, hospital accountability, and developer fault in algorithmic healthcare.

II. II. The Bolam Test, the Jacob Mathew Doctrine, and the Standard of Care

The bedrock of medical negligence jurisprudence in India is the English common law doctrine established in *Bolam v. Friern Hospital Management Committee* (1957 1 WLR 582), as definitively affirmed and modified by the Supreme Court of India in the landmark constitutional ruling of *Jacob Mathew v. State of Punjab* (2005 6 SCC 1). The *Jacob Mathew* doctrine establishes that a medical practitioner is not expected to possess the highest possible degree of professional skill; rather, the practitioner must exercise the ordinary degree of skill and competence possessed by an average member of that medical branch.

A medical professional is not negligent simply because an error of judgment occurred or because an alternative course of medical action might have yielded a superior clinical outcome. To establish liability, the claimant must prove that the doctor failed to exercise that standard of care which an ordinarily competent medical practitioner would have exercised in similar circumstances.

Applying this doctrine to clinical AI creates two distinct failure modes. The first is 'Algorithmic Abdication' (automation bias): a physician blindly accepts an AI diagnostic output without performing independent clinical verification, physical examination, or differential diagnosis. If an ordinarily competent physician would have questioned the anomalous AI recommendation, the doctor's failure to intervene constitutes actionable medical negligence. The second is 'Algorithmic Luddism': an institution fails to adopt a clinically validated AI diagnostic tool that has become the recognized standard of care, resulting in a misdiagnosis that contemporary technology would have readily prevented.

III. III. Criminal Culpability under Section 106 of Bharatiya Nyaya Sanhita 2023

Under Indian criminal law, causing death by negligence is penalized under Section 106 of the Bharatiya Nyaya Sanhita 2023 (formerly Section 304A of the Indian Penal Code 1860). Section 106(1) provides that whoever causes the death of any person by doing any rash or negligent act not amounting to culpable homicide, shall be punished with imprisonment of either description for a term which may extend to five years, and shall also be liable to fine.

However, the Supreme Court in *Jacob Mathew* established vital procedural and substantive safeguards protecting medical practitioners from unwarranted criminal harassment. The Court held that to prosecute a medical doctor under criminal law, the degree of negligence must be 'gross'—the negligence must be of such a high degree as to amount to reckless disregard for human life and safety (*mens rea*). Simple lack of care, error of judgment, or an accident is insufficient for criminal conviction.

In clinical AI contexts, a physician who relies in good faith on an approved medical device software that suffers a latent algorithmic failure cannot be convicted under Section 106 BNS. Gross criminal negligence attaches only where the practitioner deliberately ignored explicit safety alarms, deployed an uncertified research algorithm on patients without consent, or demonstrated reckless indifference to clear clinical contraindications.

IV. IV. Medical Device Regulations and Statutory Product Liability under CPA 2019

In India, clinical software used for diagnosis, prognosis, or therapeutic recommendation is legally classified as a Medical Device under the Medical Devices Rules, 2017, framed under the Drugs and Cosmetics Act, 1940. The Central Drugs Standard Control Organisation (CDSCO) categorizes medical devices into risk classes (A, B, C, and D). Clinical AI diagnostic software operating in oncology, cardiology, or intensive care fall into Class C (Moderate-High Risk) or Class D (High Risk), requiring rigorous clinical investigation, validation trials, and statutory manufacturing licenses.

Where an adverse patient outcome is caused by an intrinsic software defect—such as an unannounced model update that introduced severe classification errors—the patient is entitled to institute a statutory product liability action under Chapter VI (Sections 82-87) of the Consumer Protection Act, 2019.

Under Section 84 of the CPA 2019, the AI software manufacturer is strictly liable for design defects, deviation from manufacturing specifications, or failure to provide adequate instructions and warnings. Hospital networks and physicians can be joined as co-respondents under Section 85 (product service providers) if they failed to maintain required IT infrastructure or improperly configured the diagnostic software.

V. V. Informed Patient Consent and Algorithmic Transparency

A critical emerging requirement in healthcare AI is the doctrine of Informed Consent, codified by the Supreme Court in *Samira Kohli v. Dr. Prabha Manchanda* (2008 2 SCC 1). The Court held that a patient has an inviolable autonomy right to be informed of the nature of the diagnosis, proposed treatment, realistic risks, and viable alternatives before consenting to an intervention.

When clinical decisions are materially influenced or autonomously executed by an artificial intelligence system (such as autonomous robotic surgery), the patient must be expressly informed of the AI's involvement, its documented error rates, and the extent of human surgeon supervision. Burying AI utilization in generic hospital admission paperwork violates the *Samira Kohli* standard.

Furthermore, under the Digital Personal Data Protection Act 2023, patient health records constitute sensitive personal data requiring strict confidentiality and heightened security safeguards under Section 8(5). Training healthcare AI models on patient electronic health records (EHR) without anonymization or explicit consent exposes hospitals to severe regulatory penalties before the Data Protection Board of India.

VI. VI. Chambers Healthcare Law Practice and Medical Defense

Advocate Ritik Khurana Chambers provides specialized litigation defense and regulatory counsel to hospital conglomerates, medical device manufacturers, and healthcare AI innovators across India. Our trial practice represents medical practitioners and healthcare institutions facing medical negligence complaints before the National Consumer Commission (NCDRC), State Consumer Commissions, and State Medical Councils.

Simultaneously, our technology regulatory team advises clinical AI startups on CDSCO medical device licensing, clinical trial compliance, and ISO/IEC 42001 certification. We draft robust Institutional AI Governance Charters, Physician Supervision Guidelines, and Patient Informed Consent Protocols that reconcile cutting-edge medical technology with absolute statutory compliance.

In the delicate domain of clinical medicine, safeguarding human life requires that algorithmic innovation is anchored in clinical expertise, transparent patient communication, and unwavering legal accountability.