The Unexpected Harms of Artificial Intelligence in Healthcare: Reflections on Four Real-World Cases

Introduction: Rapid advances in Artificial Intelligence (AI), especially with large language models, present both opportunities and challenges in healthcare. This article analyzes real-world AI-related harms in healthcare. Methods: We selected four recent AI-related incidents from the AIAAIC Repository. Results: The incidents discussed include: Whisper's harmful hallucinations; UNOS's algorithm delaying transplants for black patients; the WHO's S.A.R.A.H. chatbot providing inaccurate health information; and Character AI's chatbot promoting disordered eating among teens. Discussion and conclusion: These incidents highlight diverse risks, from misinformation to safety concerns, involving both industry and institutional providers. The article emphasizes the need for systematic reporting of AI-related harms, concerns about security, privacy, and ethics, and calls for a centralized health-specific database to enhance patient safety and understanding.