公部門人工智慧應用之風險評估與內控治理──以歐盟人工智慧法及相關ISO國際標準為借鏡之法制建構

Risk Assessment and Internal Control of AI Applications in the Public Sector: Legal Framework Development with Reference to the EU AI Act and ISO Standards

隨著人工智慧技術的快速成長，全球公部門正處於數位轉型的關鍵轉折點。AI系統的導入雖能大幅提升行政效率，但其衍生的風險也帶來嚴峻挑戰。因此，如何建構一套既能促進技術創新，又能確保人權衡平且具備高度操作性的「AI風險治理架構」，已成為當前公共行政與科技法學的首要課題之一。本研究以「在地實務」、「國際指標性法規」及「國際標準」三大參考面向為核心，試圖勾勒出系統性的治理藍圖，參酌歐盟《人工智慧法》及ISO國際標準，將以風險等級為核心的分類管理邏輯，及AI專屬風險因子嵌入既有行政風險架構，並透過持續監測及「動態生命週期管理」，落實比例原則下之「必要性」與「最小侵害」要求，進而實現負責任人工智慧治理之目標。

With the rapid advancement of artificial intelligence technology, the global public sector stands at a pivotal juncture in its digital transformation. While the deployment of AI systems can significantly bolster administrative efficiency, the inherent risks present formidable challenges. Consequently, establishing an ''AI risk governance framework'' that fosters technological innovation while ensuring the protection of human rights and maintaining high operational feasibility has become a paramount issue in contemporary public administration and the legal study of technology.
This study centers on three core dimensions—local practice, international landmark regulations, and international standards—to delineate a systematic governance blueprint. By drawing upon the EU Artificial Intelligence Act and ISO international standards, the research integrates a risk-based classification logic and AI-specific risk factors into existing administrative risk management frameworks. Furthermore, through continuous monitoring and ''dynamic lifecycle management,'' it operationalizes the requirements of ''necessity'' and ''minimal impairment'' under the principle of proportionality, ultimately achieving the objective of Responsible AI Governance.