Scoping review identifies four categories of AI-related patient safety risks in nursing practice

A scoping review identified four potential categories of patient safety risk associated with artificial intelligence in nursing practice: inadequate patient assessment and safety-signal recognition, misaligned nursing recommendations and interventions, distortion of nursing documentation and handover information, and decontextualised care with weakened professional judgement. The review synthesised eight eligible studies and found that none reported direct patient harm attributable to AI use.

Published in BMC Nursing on 12 September 2026 as early-access, peer-reviewed, open-access research, the review examined evidence concerning identifiable AI applications in nursing practice or nursing workflows. The authors said AI is increasingly used in activities including care planning, assessment, documentation, handover and patient education. Eight studies met the inclusion criteria. Seven focused on large language models or other generative AI systems, and most evaluated these tools in experimental or prototype settings rather than routine clinical workflows.

The authors synthesised three categories of mechanisms associated with the reported risks: unreliable knowledge bases and model performance; misalignment with nursing work systems; and compromised information integrity and nursing representation. Safeguards described in the literature fell into three broad groups: model optimisation and context-specific validation; human review and workflow control; and governance and competency building. These proposed safeguards reflect recommendations and approaches in the included literature; the review did not establish their effectiveness in routine nursing practice.

For the review, the authors searched PubMed, Embase, CINAHL, Scopus, Web of Science, PsycINFO, IEEE Xplore and the Cochrane Library. Database searches returned 6,295 records through 28 April 2026, while supplementary searches identified 12 further records through 10 May 2026. Eligible sources described AI applications in nursing practice or workflows and reported safety risks, mechanisms or safeguards. Sources limited to non-clinical applications, nursing education or research, and technical model development were excluded, as were sources without extractable safety-relevant data.

The findings are based on a small and largely experimental evidence base, so they should not be generalised to all forms of AI or to routine nursing practice. The review authors called for evaluation of these tools in real-world or simulated nursing workflows, as well as prospective testing of safeguards that incorporate context-specific validation, nurse-led review, workflow controls and organisational governance. The supplied publisher record states that the early-access version remains subject to further editing and will be replaced by the final Version of Record.

Subject of Research: Scoping review of artificial intelligence-related patient safety risks in nursing practice

Article Title: Artificial intelligence-related patient safety risks in nursing practice: a scoping review

Article References: Zhang, P., Li, Z., She, S., Wu, H., Zheng, A., Wei, S., Ding, X., & Xiao, M. (2026). Artificial intelligence-related patient safety risks in nursing practice: a scoping review. BMC Nursing. https://doi.org/10.1186/s12912-026-05336-x

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05336-x

Keywords: artificial intelligence, patient safety, nursing practice, scoping review, generative AI, large language models, clinical workflows, human review, workflow governance, Artificial, intelligence-related, patient