Legal and Ethical Considerations of Artificial Intelligence in Medical Imaging in India
TL;DR
The ethical and legal considerations that come with using AI today are discussed, including algorithmic bias, the economic divide that limits AI access, the lack of transparency and explainability of black-box models, the threat to radiology residents from automation bias, and post-deployment model drift.
Abstract
Background Artificial intelligence is being deployed in radiology faster than the rules and regulations required to govern it. In India, this is further complicated by disparities in access to healthcare and a data protection legal framework that is in a transition phase until it becomes fully operational on May 13, 2027. Understanding these legalities is no longer optional for doctors who are in practice, using AI in the clinic, or participating in the development of AI. Methods This is a narrative review drawing on existing articles, Indian laws, and current national governance frameworks related to the use of AI in medical imaging in India. Results This review discusses the ethical and legal considerations that come with using AI today. We discuss algorithmic bias, the economic divide that limits AI access, the lack of transparency and explainability of black-box models, the threat to radiology residents from automation bias, and post-deployment model drift. On the legal side, we discuss the requirements of the Digital Personal Data Protection Act, 2023, regarding informed consent and hospitals as data fiduciaries, the classification of AI tools as Software as a Medical Device (SaMD), and Indian governance frameworks, including the Indian Council of Medical Research Ethical Guidelines for AI in Biomedical Research and Healthcare, 2023, and the Strategy for Artificial Intelligence in Healthcare for India (SAHI), 2026. Conclusion The review concludes with a practical good-practice checklist that gives radiologists an illustrative set of questions to help retain the critical mindset required at each stage of AI adoption. As AI use cases vary by context, specific guidance for your own use case is recommended.