New FMECA framework assesses patient safety risks in AI-generated clinical content

Researchers have developed and validated a new framework, based on Failure Mode, Effects, and Criticality Analysis (FMECA), to systematically assess patient safety risks associated with generative AI-created clinical content. This framework was applied to discharge summaries generated by an open LLM, GPT-OSS 120B, using real patient data from Geneva University Hospitals. The study found the framework to be usable and effective, with expert panels achieving moderate to substantial agreement in identifying failure modes and scoring their severity and detectability. AI

影响 Introduces a structured method for identifying and mitigating patient safety risks in AI-generated clinical summaries.

排序理由 This is a research paper detailing a new framework for evaluating AI safety in a specific domain. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

• FMECA
• Generative AI
• Geneva University Hospitals
• GPT-OSS 120B
• LLM

• 论文
• 安全

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →