論生成式AI時代著作權之保護與規範──從美國DABUS與Goldsmith案談起

On the AIGC Protection and Regulation in the Era of Generative AI–From DABUS to Goldsmith and Beyond

ChatGPT的橫空出世，不但引領生成式AI (generatie AI)的爆紅，也徹底改變人類與AI的互動。但大量AI生成內容(AIGC)的出現，也引發諸多著作權方面的法律爭議。過去對這些與AI有關的著作權問題，多侷限於學者的討論，但美國著作權局就DABUS與Kashtanova案的決定、其所公布涵蓋AI生成內容的著作權註冊指引、以及法院對DABUS案與Goldsmith案的判決，對AI相關的著作權問題提供了一個初步的解答。本文乃從這些案件出發，並就AI業者因應法律風險所採行的措施加以介紹，以使讀者能對生成式AI時代著作權之保護與規範有更清晰的認識與瞭解。

The emergence of ChatGPT not only led to the explosive growth of the generative AI, but also revolutionized the interaction between humans and AI. The large amount of AI generated content (AIGC), however, has also led to some basic copyright issue disputes. In the past, these AI-related copyright issues have been limited to scholarly discussions. But the U.S. Copyright Office’s decision on DABUS and Kashtanova, its promulgation of the Copyright Registration Guidelines covering AI-generated content, and the courts’decisions in DABUS and Goldsmith provide a preliminary answer to AI-related copyright issues. This article starts from these cases and the Copyright Office’s Guidelines, follows by the discussion of the measures taken by AI companies to deal with legal risks, so that readers can have a clearer understanding of the protection of copyright in the generative AI era.