Medical errors involving artificial intelligence increase hospital liability
Artificial intelligence is gradually establishing itself in the healthcare sector to support diagnostics, treatment planning, and patient communication. Hospitals already use systems based on this technology to improve early detection of certain cancers, optimize bed management, or simplify administrative tasks. However, its adoption introduces a new type of risk: medical errors related to its use.
A medical error occurs when a patient suffers unintentional harm caused by care rather than by their illness. Traditionally, these errors were primarily attributed to human factors, such as knowledge gaps or coordination failures. With the integration of artificial intelligence, errors can also occur when results generated by these systems—whether incorrect or incomplete—are applied in patient care. For example, an image analysis model might misinterpret medical scans, delaying a critical diagnosis. Similarly, some tools might downplay symptoms for specific demographic groups, negatively influencing care decisions.
These errors can have serious consequences for hospitals. They often lead to a loss of patient trust and, once publicized, damage the institution’s reputation. In the United States, nearly 17,000 medical malpractice claims are filed each year, with settlements amounting to several billion dollars. While existing research on public reactions to artificial intelligence has not yet directly studied responses to errors involving this technology, some studies suggest that people often react negatively to its use in medical practice.
A recent study explored these reactions through two experiments. The first examined how the public perceives hospital responsibility when errors originate from either a doctor, artificial intelligence, or both. Participants read a fictional scenario where a diagnostic interpretation initially deemed normal turned out to be incorrect, leading to delayed treatment and complications for the patient. The results show that participants assign more responsibility to the hospital and are more likely to consider filing a complaint or pursuing legal action when the error involves artificial intelligence, especially if it acted alone. However, when the doctor actively collaborates with the tool, negative reactions toward the hospital decrease.
The second study delved deeper into this issue by varying the types of collaboration between doctors and artificial intelligence. Participants were presented with five scenarios: a decision made solely by artificial intelligence, solely by a doctor, or through collaboration between the two, with different levels of human involvement. The results confirm that the most negative reactions occur when the initial decision relies solely on artificial intelligence. However, when the doctor examines both the areas flagged by the tool and the entire set of images, integrating this information into their own judgment, public reactions become comparable to those observed for a purely human decision.
These findings reveal that public perception depends less on the technology itself than on how it is integrated into care. Hospitals could therefore leverage this insight by emphasizing close collaboration between doctors and technological tools. Patients appear to place more trust in decisions when the involvement of healthcare professionals is visible and thorough, such as when the doctor reviews all the data rather than relying solely on the tool’s results.
This approach could help institutions benefit from the advantages of artificial intelligence while preserving patient trust in medical expertise.
References and Sources
About This Study
DOI: https://doi.org/10.1038/s44482-026-00021-x
Title: Public reactions to hospitals after adverse events involving AI
Journal: npj Digital Public Health
Publisher: Springer Science and Business Media LLC
Authors: Jungmin Choi; Yeun Joon Kim; Pengzhao Lyu; Yingyue Luna Luan; Soo Min Toh