跳到主要內容

簡易檢索 / 詳目顯示

研究生: 黃品瑄
Pin-Hsuan Huang
論文名稱: 生成式人工智慧訓練工程之著作權侵權爭議— 以創作者權益與科技發展之兼顧為主軸
指導教授: 王明禮
Ming-Li Wang
口試委員:
學位類別: 碩士
Master
系所名稱: 管理學院 - 產業經濟研究所
Graduate Institute of Industrial Economics
論文出版年: 2025
畢業學年度: 113
語文別: 中文
論文頁數: 123
中文關鍵詞: 生成式人工智慧網路爬蟲合理使用著作權
外文關鍵詞: generative artificial intelligence (GAI), web crawlers, fair use, copyright law
相關次數: 點閱:14下載:0
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  • 隨著生成式人工智慧愈發融入我們的生活當中,其訓練過程中所涉及的著作權問題,勢必將成為未來發展該技術所不可迴避的重要法律課題。關於生成式人工智慧訓練階段是否構成著作權侵權,實務上涉及對技術運作機制的理解,並須將該等技術行為具體對應至著作權侵權之構成要件。除構成要件的認定外,亦須進一步審酌是否存在抗辯事由,得以排除或限制侵權成立之可能。
    鑑於目前全球主要主導生成式人工智慧技術發展之企業多為美國公司,且從既有訴訟案件觀察,美國司法實務明顯居於主要討論地位,故本文將以美國法院之判決見解作為分析主軸,輔以歐盟數位單一市場著作權指令等重要規範,討論生成式人工智慧訓練階段是否構成著作權侵權。在侵權構成分析之外,由於當前多數生成式人工智慧開發者普遍以合理使用作為抗辯依據,本文亦將進一步探討其主張是否能主張合理使用。最後,倘若生成式人工智慧訓練行為不能主張合理使用,本文將提出具體之授權制度設計建議,以供未來建立合法資料授權架構之參考。
    儘管本文主要以美國法與美國司法實務為分析對象,惟本文所涉及之著作權概念,包括重製、暫時性重製與合理使用等,皆為各國著作權法普遍存在之基本概念,僅於適用細節上有所差異。因此,本文所提出之理論架構與分析方法,亦具一定程度之跨國參考價值,能提供理論依據與比較法之參考。


    With the rise of generative artificial intelligence (AI), the copyright issues arising from its training process are poised to become a significant legal concern. Whether the training of generative AI constitutes copyright infringement must be assessed in light of the specific operational mechanisms of such systems. In addition to determining whether infringement has occurred, it is also necessary to examine whether any defenses may apply to preclude or limit liability.
    Given that most of the leading companies developing generative AI technologies are based in the United States, and that the majority of litigation on this issue has arisen in U.S. courts, this article focuses primarily on U.S. judicial decisions as the foundation for analysis. To provide a more comprehensive view, it also draws upon relevant legal instruments, such as the European Union’s Directive on Copyright in the Digital Single Market, to explore whether the training of generative AI amounts to copyright infringement. Beyond the question of infringement, since most developers of generative AI currently rely on the doctrine of fair use as a defense, this article further examines whether such a defense can withstand judicial scrutiny. Should the fair use defense ultimately fail, this article will propose a licensing framework designed to support the lawful use of copyrighted data in the training of generative AI systems.

    目錄 摘要 I ABSTRACT II 誌謝 III 目錄 IV 1 緒論 1 1-1 研究背景與研究動機 1 1-2 研究目的與問題意識 2 1-3 研究方法 2 1-4 架構 3 1-5 用語界定 4 1-5-1 本文所指生成式 AI 4 1-5-2 本文所指稱之模型 6 2 生成式 AI 與著作權的交錯 7 2-1 生成式 AI 訓練資料現況 7 2-1-1 文本生成式 AI 7 2-1-2 圖像生成式 AI 9 2-2 生成式 AI 的訓練流程 11 2-2-1 訓練前階段 13 2-2-2 模型訓練 15 2-2-3 訓練完成後的階段 17 2-3 相關訴訟案件 18 2-4 生成式 AI 訓練的著作權侵權爭議 21 2-4-1 蒐集資料的侵權爭議 22 2-4-2 訓練模型之侵權爭議 23 2-4-3 合理使用與否 23 2-5 小結 24 3 生成式 AI 的訓練侵權爭議 27 3-1 蒐集訓練資料之侵權爭議 27 3-1-1 默示授權爭議 29 3-1-2 網頁內容可能不受著作權保護 32 3-1-3 合理使用抗辯 34 3-1-4 歐盟文字資料探勘爭議 39 3-2 模型訓練之暫時性重製侵權爭議 44 3-2-1 暫時性重製屬於著作權侵權 45 3-2-2 暫時性重製爭議之國際發展 46 3-2-3 法院所建構之暫時性重製標準 50 3-2-4 模型訓練難以被認定為暫時性重製 53 3-3 小結 55 4 訓練生成式 AI 的合理使用 59 4-1 著作權之限制 59 4-2 訓練生成式 AI 的合理使用分析 60 4-2-1 第一因素:利用的目的和性質 60 4-2-2 第四因素:使用對受著作權保護之作品潛在市場或價值之影響 69 4-2-3 其他因素 76 4-3 THOMSON REUTERS ENTERPRISE CENTRE GMBH V. ROSS INTELLIGENCE INC.案 81 4-4 生成式 AI 的訓練使用並非合理使用 83 5 建立授權制度之探討 89 5-1 執行授權制度必須克服之困難 89 5-1-1 成本過高 89 5-1-2 合理公平取得訓練資料 89 5-2 可能的授權制度 90 5-2-1 集體管理組織 90 5-2-2 強制授權 91 5-2-3 擴大集體授權制度 92 5-2-4 其他可能授權方法 93 5-3 授權制度分析 95 5-3-1 集體管理組織 95 5-3-2 強制授權 96 5-3-3 擴大集體授權制度 98 5-4 回歸授權制度所帶來的面影響 99 5-4-1 訓練資料品質提升 99 5-4-2 可促進生成式 AI 訓練資料透明化 100 5-4-3 緩解生成式 AI 自行監督壓力 101 5-5 小結 104 6 結論 107 參考文獻 110

    中文參考文獻
    林宜柔、許正乾、陳家駿,AI/ChatGPT v.智慧財產權──美國生成式AI案例評析,1版,元照,頁42-47,2024年。
    陳家駿,AI人工智能vs智慧財產權,2版,元照,頁2-4,2022年。
    章忠信,美國一九九八年數位化千禧年著作權法案簡介,萬國法律,第107期,頁34,1999年。

    英文參考文獻
    Amarikwa, Melany, Internet Openness at Risk: Generative Ai's Impact on Data Scraping, 30 Rich. J.L. & Tech, 533 (2024).
    Art, Neill & Thomas, James & Lee, Erika, A Framework for Applying Copyright Law to The Training of Textual Generative Artificial Intelligence, 32 TEX. INTELL. PROP. L.J. 225 (2024).
    Asay, Clark D. & Sloan, Arielle & Sobczak, Dean, Is Transformative Use Eating the World, 61 B.C. L. REV. 905 (2020).
    Band, Jonathan & Marcinko, Jeny, A New Perspective on Temporary Copies: The Fourth Circuit's Opinion in Costar V. Loopnet, 2005 STAN. TECH. L. REV. P1 (2005).
    Bogden, Melissa A., Fixing Fixation: The Ram Copy Doctrine, 43 ARIZ. ST. L.J. 181 (2011).
    Carroll, Michael W., Copyright and The Progress of Science Why Text and Data Mining Is Lawful, 53 U.C. DAVIS L. REV. 893 (2019).
    Casey, Bryan & Lemley, Mark A., Fair Learning, 99 TEX. L. REV. 743 (2021).
    Cooper, A. Feder & Grimmelmann, James & Lee, Katherine, Talkin’ ’Bout AI Generation: Copyright and the Generative-AI Supply Chain, CSLAW '24: PROCEEDINGS OF THE SYMPOSIUM ON COMPUTER SCIENCE AND LAW (2024), at https://dl.acm.org/doi/pdf/10.1145/3614407.3643696.
    Culliton, Brianne M., The Generative AI Pirate? The Intersection of Copyrights and Generative AI In Literary Works, 15 CYBARIS AN INTELL. PROP. L. REV. 251 (2024).
    Gal, Michal S. & Lynskey, Orla, Synthetic Data: Legal Implications of The Data-Generation Revolution, 109 IOWA L. REV. 1087 (2024).
    Ginsburg, Jane C., Fair Use in the US Redux: Reformed or Still Deformed, 2024 SING. J. LEGAL STUD. 52 (2024).
    Hashimoto, Tatsunori & Henderson, Peter & Jurafsky, Dan & Lemley, Mark A. & Li, Xuechen & Liang, Percy, Foundation Models And Fair Use, JOURNAL OF MACHINE LEARNING RESEARCH, Sep. 23, 2023, at https://www.jmlr.org/papers/volume24/23-0569/23-0569.pdf .
    Hayes, David L., Advanced Copyright Issues on the Internet, 7 TEX. INTELL. PROP. L.J. 1, 7 (1998).
    Lemley, Mark A., Property, Intellectual Property, And Free Riding, 83 TEX. L. REV. 1031 (2005).
    Leval, Pierre N., Toward A Fair Use Standard, 103 HARV. L. REV. 1105 (1990).
    Library of Congress Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence (Mar. 16, 2023), at https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence.
    Maffioli, Daniel Rodriguez, Copyright in Generative AI training: Balancing Fair Use through Standardization and Transparency (Aug.21, 2023), at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4579322.
    Mantegna, Micaela, Artificial: Why Copyright Is Not the Right Policy Tool to Deal with Generative AI, 133 YALE L.J. FORUM 1126 (2024).
    Murray, Michael D., Generative AI Art: Copyright Infringement and Fair Use, 26 SMU SCI. & TECH. L. REV. 259 (2023).
    Najork, Marc & Olston, Christopher, Web Crawling, 4 FOUND. & TRENDS IN INFO. RETRIEVAL 175 (2010), at http://i.stanford.edu/~olston/publications/crawling_survey.pdf.
    Nilsson, Nils J., THE QUEST FOR ARTIfiCIAL INTELLIGENCE: A HISTORY OF IDEAS AND ACHIEVEMENTS (2010).
    Opderbeck, David W., Copyright in AI Training Data: A Human-Centered Approach, 76 OKLA. L. REV. 951 (2024).
    Quang, Jenny, Does Training Ai Violate Copyright Law, 36 BERKELEY TECH. L.J. 1407 (2021).
    Rattzén, Mattias, Location Is All You Need: Copyright Extraterritoriality and Where to Train Your AI, 26 COLUM. SCI. & TECH. L. REV. 175 (2024).
    Sableman, Mark, Link Law Revisited: Internet Linking Law at Five Years, 16 BERKELEY TECH. L.J. 1273 (2001).
    Sag, Matthew, Copyright Safety for Generative AI, 61 HOUS. L. REV. 295 (2023).
    Samuelson, Pamela, Generative AI Meets Copyright, SCIENCE VOL. 381, NO. 6654.
    Shipley, David E., A Transformative Use Taxonomy: Making Sense of the Transformative Use Standard, 63 WAYNE L. REV. 267 (2018).
    Sobel, Benjamin L. W., Artificial Intelligence's Fair Use Crisis, 41 COLUM. J.L. & ARTS 45 (2017).
    Spica, Elizabeth, Public Interest, The True Soul: Copyright's Fair Use Doctrine and the Use of Copyrighted Works to Train Generative AI Tools, 33 TEX. INTELL. PROP. L.J. 67 (2024).
    Subia Espinoza, Danna, The Future of Art and Copyright in the World of AI, 32 CATH. U. J. L. & TECH. 189 (2024).
    Torrance, Andrew W. & Tomlinson, Bill, Training Is Everything: Artificial Intelligence, Copyright, and “Fair Training”, 128 DICK. L. REV. 233 (2023).

    QR CODE
    :::