July 2025
U.S. Copyright Office’s Three-Part Report on Copyright and AI Part II: The Copyrightability of AI-Generated Content
From July 2024 through May 2025, the U.S. Copyright Office released a trilogy of reports addressing issues at the intersection of generative artificial intelligence (AI) and copyright law. These reports cover: (1) Digital Replicas, (2) Copyrightability, and (3) Training of Generative AI. Having introduced the first report
[1]
previously, this article continues with an overview of the second report’s content and recommendations, followed by a discussion of its implications for Taiwan.
Overview of the Technological Background of Generative AI
To address questions regarding authorship and ownership of AI-generated content, it is first necessary to understand the technological background underpinning AI systems. Generative AI refers to systems that use machine learning models (such as large language models) trained on vast datasets to generate content including text, images, and music. Users input prompts to communicate the desired characteristics of the output, and the AI responds based on its training data to generate a corresponding result.
Judicial Standards for Copyrightability in the United States
Drawing from landmark U.S. cases such as Feist [2] , Burrow-Giles [3] , and Andrien [4] , the prevailing judicial standards for copyrightability can be summarized as follows: Copyright protection is available only for works that are “original,” meaning they must be the product of human authorship and demonstrate at least a minimal degree of creativity. Moreover, even when a work is produced through a camera or other mechanical device, if the creator has exercised genuine creative input, the work may still qualify for copyright protection.
In Andrien, the court defined an “author” as “party who actually creates the work, that is, the person who translates an idea into an expression that is embodied in a copy by himself or herself, or who authorizes another to embody the expression in a copy.” The issue of AI and authorship builds upon this foundational concept and extends it further.
Are Outputs Generated from Prompts Input into AI Considered Human Authorship?
The use of prompts serves as a bridge for communication between humans and AI systems. Yet, whether the resulting outputs generated through such prompts constitute “human authorship” and thus qualify as copyrightable remains a central question.
1. Summary of Public Comments Solicited by the U.S. Copyright Office
A majority of respondents maintained a negative view, asserting that the mere act of inputting prompts is insufficient for a user to claim copyright in the resulting AI-generated content.
The Authors Guild emphasized the “unpredictability” inherent in the relationship between prompts and resulting outputs, arguing that users lack the necessary control over the output to demonstrate a sufficient closeness between “conception” and “execution.” Adobe likewise stated that although users may provide specific prompts to an AI system, the resulting output is entirely dependent on the AI’s interpretation of those prompts—an interpretation that, in Adobe’s view, does not qualify for copyright protection.
2. Users Cannot Generally Obtain Copyright Simply by Inputting Prompts
The Copyright Office clearly stated in its report that the act of inputting prompts is essentially a way of conveying instructions or ideas. No matter how detailed or specific the prompts may be, they are generally insufficient to establish the user as the author of the AI-generated output. This is primarily because users do not control how the AI processes and generates its output.
According to a standard articulated by the Third Circuit, a party commissioning a work may be considered the author only when the execution is purely mechanical or routine, without substantive intellectual contribution or technical enhancement by the executing party. While entering prompts into an AI system may resemble such mechanical execution, the Copyright Office noted a critical distinction. In human collaborations, the commissioning party typically retains the ability to oversee and direct the process to ensure alignment with the original intent. In contrast, users lack enough control over how the AI interprets and processes the prompts. In other words, while a prompt may reflect the user’s internal ideas or intentions, the user cannot dictate how the AI interprets or realizes those ideas.
Limited Cases Eligible for Copyright Protection
1. Expressive Inputs
Expressive inputs refer to user-provided materials that embody original expression and concretely reflect human creative intent, as opposed to mere instructive prompts. When AI-generated works are meaningfully derived from such expressive inputs and reflect original human authorship, they may qualify for copyright protection.
A representative example includes a user providing an original image via prompt and supplementing it with additional instructions. In such cases, the user may exert a relatively higher degree of control and be able to reasonably anticipate the nature of the AI-generated output. The output may also reflect distinct features of the original human-authored material. Accordingly, copyright protection may apply to: (1) the human-authored original component; and (2) the creative “selection, coordination, and arrangement” of the original and AI-generated elements.
2. Modifying or Arranging AI-Generated Content
Where a human modifies AI-generated material and the modification meets the threshold of creativity required for copyright protection, the resulting work may qualify for copyright. In addition, the Copyright Office’s “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence [5] ” affirms that a human’s creative selection or arrangement of AI-generated content may also be eligible for protection under existing copyright law.
The U.S. Copyright Office’s Position on AI-Generated Content and Copyright Protection
Based on its analysis of existing copyright law and policy, and in light of the public comments received in response to its Notice of Inquiry (NOI), the U.S. Copyright Office has drawn several conclusions and recommendations. First, with respect to the copyrightability of AI-generated content, the Office concludes that existing law is sufficient and that legislative change is not currently warranted. Second, when AI tools are used merely to assist human creativity, such use does not preclude copyright protection for the resulting work. So long as the work contains original expression authored by a human, that portion of the work remains eligible for copyright, even if it also includes material generated by AI.
By contrast, content that is entirely generated by AI, or in which the human lacks sufficient control over the expressive elements, falls outside the scope of copyright protection. Whether a human’s contribution to an AI-generated output is sufficient to confer authorship must be determined on a case-by-case basis. Under currently available technologies, the mere act of inputting prompts generally does not provide the degree of control necessary for a human to be considered the author of the resulting content.
However, where AI-generated output contains human-authored expression that is both sufficiently original and perceptible, or where the human has exercised creative judgment in selecting, organizing, arranging, or modifying AI-generated material, that portion may be eligible for copyright protection as a human-authored work. The Office further notes that, at this stage, there is no need for additional copyright or sui generis protection for AI-generated content.
The Office will continue to monitor technological and legal developments, reassess its position as appropriate, and provide the public with updated guidance and resources.
Implications for Taiwan
In Taiwan, the legal treatment of AI-generated content under copyright law remains in a developmental stage. According to guidance issued by the Taiwan Intellectual Property Office (TIPO) on May 22, 2025 [6] , when AI is used merely as an auxiliary tool—such as illustration software—and the work includes creative contributions from a human author, the resulting work is, in principle, eligible for copyright protection. Conversely, where the content is entirely generated by AI without any human creative input, it does not qualify for copyright protection.
Nonetheless, questions concerning whether an AI-generated image constitutes a copyrightable work or infringes upon the rights of others must be evaluated by judicial authorities on a case-by-case basis, based on the specific facts and evidence presented. Given the diversity of disputes arising from this issue and the current lack of consistent legal standards in practice, Taiwan’s administrative and judicial bodies may look to the U.S. Copyright Office’s approach for reference. In doing so, they may work toward establishing a balanced framework that both protects copyright and encourages creativity—one that can serve as a foundation for the future evolution of Taiwan’s copyright system.
[1] Leetsai. (2024, May 15). U.S. Copyright Office’s three-part report on copyright and AI—Part I: Challenges and legal responses to digital replicas. Leetsai. https://www.leetsai.com/u-s-copyright-offices-three-part-report-on-copyright-and-aipart-i-challenges-and-legal-responses-to-digital-replicas?lang=en-US
[2] Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 344-45 (1991)
[3] Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 55–57 (1884)
[4] Andrien v. Southern Ocean County Chamber of Commerce, 927 F.2d 132, 134–35 (3d Cir. 1991)
[5] Federal Register (2023, March 16). Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence. https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence
[6] Intellectual Property Office, Ministry of Economic Affairs, Email 1140522c, https://www.tipo.gov.tw/tw/copyright/692-34252.html?highlight=
Overview of the Technological Background of Generative AI
To address questions regarding authorship and ownership of AI-generated content, it is first necessary to understand the technological background underpinning AI systems. Generative AI refers to systems that use machine learning models (such as large language models) trained on vast datasets to generate content including text, images, and music. Users input prompts to communicate the desired characteristics of the output, and the AI responds based on its training data to generate a corresponding result.
Judicial Standards for Copyrightability in the United States
Drawing from landmark U.S. cases such as Feist [2] , Burrow-Giles [3] , and Andrien [4] , the prevailing judicial standards for copyrightability can be summarized as follows: Copyright protection is available only for works that are “original,” meaning they must be the product of human authorship and demonstrate at least a minimal degree of creativity. Moreover, even when a work is produced through a camera or other mechanical device, if the creator has exercised genuine creative input, the work may still qualify for copyright protection.
In Andrien, the court defined an “author” as “party who actually creates the work, that is, the person who translates an idea into an expression that is embodied in a copy by himself or herself, or who authorizes another to embody the expression in a copy.” The issue of AI and authorship builds upon this foundational concept and extends it further.
Are Outputs Generated from Prompts Input into AI Considered Human Authorship?
The use of prompts serves as a bridge for communication between humans and AI systems. Yet, whether the resulting outputs generated through such prompts constitute “human authorship” and thus qualify as copyrightable remains a central question.
1. Summary of Public Comments Solicited by the U.S. Copyright Office
A majority of respondents maintained a negative view, asserting that the mere act of inputting prompts is insufficient for a user to claim copyright in the resulting AI-generated content.
The Authors Guild emphasized the “unpredictability” inherent in the relationship between prompts and resulting outputs, arguing that users lack the necessary control over the output to demonstrate a sufficient closeness between “conception” and “execution.” Adobe likewise stated that although users may provide specific prompts to an AI system, the resulting output is entirely dependent on the AI’s interpretation of those prompts—an interpretation that, in Adobe’s view, does not qualify for copyright protection.
2. Users Cannot Generally Obtain Copyright Simply by Inputting Prompts
The Copyright Office clearly stated in its report that the act of inputting prompts is essentially a way of conveying instructions or ideas. No matter how detailed or specific the prompts may be, they are generally insufficient to establish the user as the author of the AI-generated output. This is primarily because users do not control how the AI processes and generates its output.
According to a standard articulated by the Third Circuit, a party commissioning a work may be considered the author only when the execution is purely mechanical or routine, without substantive intellectual contribution or technical enhancement by the executing party. While entering prompts into an AI system may resemble such mechanical execution, the Copyright Office noted a critical distinction. In human collaborations, the commissioning party typically retains the ability to oversee and direct the process to ensure alignment with the original intent. In contrast, users lack enough control over how the AI interprets and processes the prompts. In other words, while a prompt may reflect the user’s internal ideas or intentions, the user cannot dictate how the AI interprets or realizes those ideas.
Limited Cases Eligible for Copyright Protection
1. Expressive Inputs
Expressive inputs refer to user-provided materials that embody original expression and concretely reflect human creative intent, as opposed to mere instructive prompts. When AI-generated works are meaningfully derived from such expressive inputs and reflect original human authorship, they may qualify for copyright protection.
A representative example includes a user providing an original image via prompt and supplementing it with additional instructions. In such cases, the user may exert a relatively higher degree of control and be able to reasonably anticipate the nature of the AI-generated output. The output may also reflect distinct features of the original human-authored material. Accordingly, copyright protection may apply to: (1) the human-authored original component; and (2) the creative “selection, coordination, and arrangement” of the original and AI-generated elements.
2. Modifying or Arranging AI-Generated Content
Where a human modifies AI-generated material and the modification meets the threshold of creativity required for copyright protection, the resulting work may qualify for copyright. In addition, the Copyright Office’s “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence [5] ” affirms that a human’s creative selection or arrangement of AI-generated content may also be eligible for protection under existing copyright law.
The U.S. Copyright Office’s Position on AI-Generated Content and Copyright Protection
Based on its analysis of existing copyright law and policy, and in light of the public comments received in response to its Notice of Inquiry (NOI), the U.S. Copyright Office has drawn several conclusions and recommendations. First, with respect to the copyrightability of AI-generated content, the Office concludes that existing law is sufficient and that legislative change is not currently warranted. Second, when AI tools are used merely to assist human creativity, such use does not preclude copyright protection for the resulting work. So long as the work contains original expression authored by a human, that portion of the work remains eligible for copyright, even if it also includes material generated by AI.
By contrast, content that is entirely generated by AI, or in which the human lacks sufficient control over the expressive elements, falls outside the scope of copyright protection. Whether a human’s contribution to an AI-generated output is sufficient to confer authorship must be determined on a case-by-case basis. Under currently available technologies, the mere act of inputting prompts generally does not provide the degree of control necessary for a human to be considered the author of the resulting content.
However, where AI-generated output contains human-authored expression that is both sufficiently original and perceptible, or where the human has exercised creative judgment in selecting, organizing, arranging, or modifying AI-generated material, that portion may be eligible for copyright protection as a human-authored work. The Office further notes that, at this stage, there is no need for additional copyright or sui generis protection for AI-generated content.
The Office will continue to monitor technological and legal developments, reassess its position as appropriate, and provide the public with updated guidance and resources.
Implications for Taiwan
In Taiwan, the legal treatment of AI-generated content under copyright law remains in a developmental stage. According to guidance issued by the Taiwan Intellectual Property Office (TIPO) on May 22, 2025 [6] , when AI is used merely as an auxiliary tool—such as illustration software—and the work includes creative contributions from a human author, the resulting work is, in principle, eligible for copyright protection. Conversely, where the content is entirely generated by AI without any human creative input, it does not qualify for copyright protection.
Nonetheless, questions concerning whether an AI-generated image constitutes a copyrightable work or infringes upon the rights of others must be evaluated by judicial authorities on a case-by-case basis, based on the specific facts and evidence presented. Given the diversity of disputes arising from this issue and the current lack of consistent legal standards in practice, Taiwan’s administrative and judicial bodies may look to the U.S. Copyright Office’s approach for reference. In doing so, they may work toward establishing a balanced framework that both protects copyright and encourages creativity—one that can serve as a foundation for the future evolution of Taiwan’s copyright system.
[1] Leetsai. (2024, May 15). U.S. Copyright Office’s three-part report on copyright and AI—Part I: Challenges and legal responses to digital replicas. Leetsai. https://www.leetsai.com/u-s-copyright-offices-three-part-report-on-copyright-and-aipart-i-challenges-and-legal-responses-to-digital-replicas?lang=en-US
[2] Feist Publ’ns, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340, 344-45 (1991)
[3] Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 55–57 (1884)
[4] Andrien v. Southern Ocean County Chamber of Commerce, 927 F.2d 132, 134–35 (3d Cir. 1991)
[5] Federal Register (2023, March 16). Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence. https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence
[6] Intellectual Property Office, Ministry of Economic Affairs, Email 1140522c, https://www.tipo.gov.tw/tw/copyright/692-34252.html?highlight=
