生成式AI於醫療領域應用之管制議題

Regulatory Issues for the Application of Generative AI in Healthcare

生成式AI可生成文字、圖像、影片、音樂、電腦程式碼或相關組合等內容，其利用深度學習技術，得識別和分類訓練資料集中的模式、風格或結構。本文以生成式AI於醫療領域應用下衍生的問題為核心，嘗試討論當中可能產生的法律規範衝擊與影響，並且一併說明歐盟人工智慧法草案中關於生成式AI之規範，探討此等管制架構對醫療使用生成式AI的影響程度為何。生成式AI管制重點之一在於使系統具備可信賴性，避免偏見歧視或是虛假資料流傳，而在醫療領域應用的同時，更應考量病人的隱私與安全性，避免醫療事故產生，因此，除了強調依不同風險程度管制與遵守透明性義務以外，並應有相對應的驗證機制檢視是否合於規範。

Generative AI is a type of artificial intelligence that can generate new content such as text, images, videos, music, computer code, or related combinations. It relies on deep learning that can identify and classify patterns or structure within the training datasets. This paper explores some major application of generative AI in the medicine and healthcare and attempts to discuss some associated legal issues, such as safety, reliability, privacy, and transparency obligations. The EU draft Artificial Intelligence Act will also be discussed in this paper to know how it establishes a comprehensive legal framework for developing and using generative AI. The core principle of AI regulation is to ensure the trustworthy of the AI system and avoid bias and discrimination and the spread of fake information. Therefore, in addition to introducing the transparency obligations based on different levels of risk, there should be an AI verification mechanism as a tool to ensure regulatory compliance.