初探ChatGPT管制挑戰與回應──美國法的觀點

An Overview of the Regulatory Challenges and Responses of ChatGPT in the US Law

ChatGPT的出現，使得人工智慧的應用再度受到公眾關注，而其模型與訓練資料龐大、泛用於不同用途之通用性、立基於過去資料進行預測、內容可信度問題等特殊性，帶來新的管制挑戰。相較於歐盟法上對於人工智慧逐步進行立法，美國聯邦法向來代表的是不同管制模式，對於人工智慧的管制採取部門式規範及軟性規範，而ChatGPT的風險在美國引發科學與公民社群的重視，但政治上尚未達成共識，僅有形成專業自治規範。在特定議題如個人資料保護、不實內容的責任上，現行法的解釋適用亦有所侷限，亦凸顯管制因應之重要。本文爬梳美國法律與政策對於人工智慧以及ChatGPT的管制發展與爭議，以作為我國之參考。

ChatGPT leads a new wave of public attention to artificial intelligence (AI) development. The characteristics of ChatGPT, namely the large scale of models and training data, the generality applicable for different purposes, the ability of prediction based on past data, and its credibility to generate content, cause new regulatory challenges. Compared with the EU law that gradually legislates over AI, US federal law has been representing a different paradigm, which takes the sectoral and soft regulations approach. Despite the scientific and civil community’s robust debates over the risks of ChatGPT, the legislation of this emerging technology does not reach political consensus but can only form professional self-regulation. Legal issues raised by ChatGPT, such as personal data protection or the liability of false content, can reveal the current law’s limit and the necessity of reconsidering it. This article reviews the US AI law and policy and the regulatory challenges and responses brought by ChatGPT, which could also serve as a reference to our ongoing discussion over forming our domestic AI regulatory framework.