The ethical challenge of AI in medicine: Who owns the mistakes?

Letter

To the Editor, Artificial intelligence encompasses machine learning and natural language processing to simulate human cognition. It is revolutionizing medicine by supporting diagnosis and data management in healthcare. Its advantages include shorter hospital stays and reduced use of invasive procedures, which ultimately lower treatment costs, increase efficiency, and improve patient outcomes and quality of life. Despite these benefits, AI may produce errors with serious consequences, highlighting challenges regarding accountability when issues such as responsibility and liability arise in clinical practice. The application of artificial intelligence in medical fields creates serious ethical, social, and legal challenges, which include concerns regarding patient privacy and data security, algorithmic bias, and fair access to AI-powered medical treatment. The existing problems demonstrate the requirement for strong regulatory systems and explicit responsibility guidelines, which will guarantee that medical institutions use artificial intelligence technologies in a responsible manner. ¹

Currently, AI applications are being utilized clinically for disease screening and triage, diagnosis, risk assessment, and treatment planning. Notable examples include the IDx-DR diagnostic system for diabetic retinopathy screening, the CC-Cruiser for childhood cataract detection, and MySurgeryRisk for preoperative risk evaluation. Literature also highlights AI applications in cancer, neurological disorders, eye diseases, infections, and musculoskeletal conditions. These tools support clinical decision-making, enhance diagnostic efficiency and accuracy, and are reshaping modern healthcare delivery. ²

Despite these benefits, there is a pressing need to improve these systems due to various legal and ethical concerns, such as built-in biases, a lack of privacy and transparency, data leaks, and uncertain accuracy, all of which should be thoroughly validated and continuously evaluated. Legal and ethical frameworks for artificial intelligence in healthcare currently face development challenges, which lead to multiple uncertainties regarding liability requirements. Naik et al. demonstrate that AI system errors create complicated problems because they require determination of which party should take responsibility; a determination that involves deciding whether liability falls on the healthcare provider using the tool, the hospital or institution deploying it, or the technology developers and companies that designed the algorithm. The existing gaps require identification of specific guidelines together with regulatory frameworks that will protect patient safety while allowing for innovative solutions to emerge. ³ The increasing use of AI in diagnostics and treatment has raised the question: when AI commits an error, who should be held accountable, the doctor, the institution, or the algorithm developer?

A clear legal framework and ethical guidelines are required for the safe and reliable use of AI, and responsibility for errors should be shared among the designer, manufacturer, and the user of the system. ⁴ Patient welfare should remain central when designing and calibrating AI so that the system is not only accurate and transparent but also accountable. ⁵ A robust regulatory and implementation model will promote greater accountability, ultimately enhancing patients’ trust and the safety of healthcare delivery.

In conclusion, if an error arises from flaws within the algorithm, responsibility should be shared. The physician, as the ultimate decision-maker, along with the designers and manufacturers, must recognize that any error can seriously harm patients. Therefore, a collaborative, interdisciplinary approach is essential for the safe design and implementation of AI systems.