Artificial Intelligence Literacy in Intensive Care: From Algorithmic Fluency to Clinical Accountability

Background: Artificial intelligence (AI), particularly large language models (LLMs), is increasingly being used in intensive care environments to assist with documentation, literature synthesis, and clinical information retrieval.While these tools offer efficiency and cognitive support in high-pressure settings such as the intensive care unit (ICU), they also introduce risks related to hallucination, automation bias, and a lack of explainability.Aim: To examine the emerging challenges associated with AI use in critical care-especially hallucination and algorithmic overconfidence-and to highlight the need for structured AI literacy and governance frameworks for intensivists.Findings: Recent analyses show that AI hallucination-the generation of fabricated or distorted information-is an intrinsic feature of probabilistic language models rather than a rare technical error.In critical care contexts, this may manifest as fabricated references, distorted clinical reasoning, or overconfident recommendations.Additional concerns include automation bias, where clinicians may inadvertently rely on algorithmic outputs, and opacity in AI decision-making processes.Emerging evidence demonstrates that while AI tools may enhance efficiency in tasks such as documentation and knowledge summarization, they remain unreliable for unsupervised clinical reasoning. Conclusion:The integration of AI into intensive care practice should be guided by structured AI literacy, institutional governance frameworks, and continued human oversight.Artificial intelligence can serve as a supportive cognitive tool, but responsibility for clinical judgment, patient safety, and ethical decision-making must remain firmly with the clinician.