Ethical Risks of Medical AI Agents and a Governance Framework: A Systematic Review of Current Applications and Policy Responses
Health Care Science · 17.11.2026
DOI: 10.1002/hcs2.70111 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift Health Care Science, online erschienen am 17.11.2026. Verfasst von Yafang Li und 7 weiteren Autor:innen. Die Arbeit nennt 75 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This systematic review identified ethical risks associated with medical AI agents and developed a governance framework adapted to China's healthcare context. PubMed, Web of Science, China National Knowledge Infrastructure, and Wanfang Data were searched for records published or publicly released between January 1, 2020 and 30 June 2025. Of 1239 records identified, 837 underwent title and abstract screening, 68 were assessed in full, and 58 were included in the synthesis. The evidence base comprised 12 core agentic sources, 24 agent‐focused boundary sources, and 22 governance, policy, legal, or normative sources. Eight non‐exclusive ethical‐risk themes were identified, led by safety, reliability, and hallucination; accountability and liability; bias and fairness; and privacy and security. Core agentic evidence indicated that multi‐step execution, external tool use, persistent memory, and multi‐agent coordination may amplify error propagation, provenance, privacy, and responsibility‐allocation risks. Six representative cases and governance events were analyzed separately using the Four Principles of Biomedical Ethics to examine how similar risks arise across different technologies, jurisdictions, and institutional settings. The synthesis informed a four‐component governance framework linking ethical and legal challenges, biomedical ethical principles, operational governance mechanisms, and expected outcomes. The proposed mechanisms include lifecycle risk monitoring, human oversight and escalation, responsibility allocation, privacy and data governance, documentation and auditability, fairness and equitable‐access safeguards, and multistakeholder ethics review. The framework is intended as a policy‐oriented and institutionally adaptable proposal rather than a validated intervention and requires prospective evaluation in specific cases of healthcare organizations and governance events.
Zitieren
APA
Li, Y., Zhao, L., Yang, Y., Xin, T., Bian, Z., Chen, Y., Zhang, Z., & Zhang, Z. (2026). Ethical Risks of Medical AI Agents and a Governance Framework: A Systematic Review of Current Applications and Policy Responses. Health Care Science. https://doi.org/10.1002/hcs2.70111
Vancouver
Li Y, Zhao L, Yang Y, Xin T, Bian Z, Chen Y, et al. Ethical Risks of Medical AI Agents and a Governance Framework: A Systematic Review of Current Applications and Policy Responses. Health Care Science. 2026. doi:10.1002/hcs2.70111
BibTeX
@article{Li2026,
author = {Li, Yafang and Zhao, Lina and Yang, Yiyi and Xin, Tangsiwei and Bian, Zihao and Chen, Yiyang and Zhang, Zeyu and Zhang, Zongjiu},
title = {Ethical Risks of Medical AI Agents and a Governance Framework: A Systematic Review of Current Applications and Policy Responses},
journal = {Health Care Science},
year = {2026},
doi = {10.1002/hcs2.70111}
}
EUVIMED