Seven deadly sins in artificial intelligence for digital medicine

Area of Science:

• Medical Ethics
• Artificial Intelligence
• Clinical Informatics

Background:

• Artificial intelligence (AI) integration in clinical settings is rapidly increasing.
• This rapid adoption raises significant ethical concerns regarding trust, fairness, empathy, and governance.
• The ethical landscape for AI in medicine is currently unstable and lacks comprehensive frameworks.

Purpose of the Study:

• To introduce a novel conceptual framework, the "Seven Deadly Sins of AI in Medicine," detailing recurring systemic failure modes.
• To validate this framework through a global survey of healthcare stakeholders.
• To propose actionable principles, the "Seven Cardinal Virtues," for responsible AI development and governance in medicine.

Main Methods:

• Systematic synthesis of scientific literature, clinical guidelines, and regulatory frameworks to develop the initial framework.
• A global, cross-professional opinion poll of 914 stakeholders from 143 countries (July 2024 - March 2025) to validate the pre-established framework.
• Analysis of survey results to identify areas of agreement and divergence in ethical concerns and regulatory attitudes.

Main Results:

• The "Seven Deadly Sins" framework, encompassing Blind Trust, Overregulation, Dehumanization, Misaligned Optimization, Overinforming/False Forecasting, Misapplied Statistics, and Self-Referential Evaluation, was confirmed by global stakeholders.
• Broad agreement was found across diverse professional backgrounds and cultures regarding the identified AI risks in medicine.
• Significant divides in attitudes toward regulation were observed, particularly between technologically advanced nations and emerging economies.

Conclusions:

• The "Seven Deadly Sins" framework serves as a diagnostic tool for identifying ethical risks in medical AI.
• The proposed "Seven Cardinal Virtues" offer a pathway toward trustworthy and human-centered AI in healthcare.
• A unified approach is needed to move beyond scattered ethical guidelines for effective AI governance in medicine.