June 2025
U.S. Copyright Office’s Three-Part Report on Copyright and AI
Part I: Challenges and Legal Responses to Digital Replicas
As generative artificial intelligence (AI) becomes widely used, legal and policy disputes over copyright are intensifying. In response, the U.S. Copyright Office released a three-part report based on over 10,000 public comments submitted after its August 2023 Notice of Inquiry:
1. Part1: Digital Replicas (July 2024) 1
2. Part2: Copyrightability (January 2025) 2
3. Part3: Generative AI Training (May 2025) 3
These reports aim to guide Congress, courts, and executive agencies toward a balanced approach—advancing technology while protecting human creativity. This summary highlights the first report on Digital Replicas and its implications for Taiwan.
Definition and Technological Context
“Digital replicas” refer to highly realistic reproductions of a person’s voice, likeness, or image using digital tools, often referred to as Deepfakes. With rapid AI advances, such content can now be created swiftly, convincingly, and without consent—raising new challenges for copyright, personality rights, and social trust.
Case Spotlight: Scarlett Johansson vs. OpenAI
In May 2024, OpenAI launched a real-time voice interaction feature for ChatGPT. One of the developed voices, “Sky,” drew public attention for its striking resemblance to the voice performance of renowned actress Scarlett Johansson in the film “Her.” Johansson expressed shock and anger and, through her attorney, formally sent a letter to OpenAI demanding: “Take down my voice immediately!” This incident highlights the potential personality rights issues that may arise from the commercial use of digital replica technologies. 4
Social Impact and Legal Gaps
Beyond the frequent targeting of celebrities and political figures, what alarms the public even more is the cloning of ordinary individuals' voices or images to deceive their friends and family in scams. This means anyone could become a victim of unauthorized digital replicas, posing a serious threat to the foundation of social trust and highlighting the urgent need for updates to the current legal framework.
Analysis of U.S. Legal Framework
1. The U.S. Copyright Office observed that public commentary overwhelmingly reflects a shared concern: existing U.S. laws are ill-equipped to govern the rising tide of unauthorized digital replicas. There is a pressing need for new federal legislation to meet the moment and respond to this rapidly evolving challenge.
The inadequacies of the existing U.S. legal framework and its regulations include:
(1) State Law
In the United States, most state laws regulate unauthorized digital replicas primarily through the Right of Privacy and the Right of Publicity. The Right of Privacy is intended to protect individuals from unreasonable intrusions, including false representations or the misappropriation of personal identity. Relevant torts under this right include Appropriation and False Light. The Right of Publicity, on the other hand, focuses on the use of personal identity elements—such as name, image, or voice—for commercial purposes, in order to safeguard their economic value. In recent years, some states have begun to enact or amend laws in response to this issue. For instance, Tennessee has extended the right of publicity to include protection for voice imitation, expanding its coverage even to non-commercial dissemination.
However, state-level regulations remain notably inadequate in addressing unauthorized digital replicas. The main challenge lies in the lack of consistency across states: laws differ in terms of who is protected (e.g., only celebrities or all individuals), whether protection applies only to commercial uses, and whether rights continue after death, among other variables.
(2) Federal Law
Although there is currently no specific federal law in the United States that directly regulates digital replicas, several existing laws may still apply.
a. Copyright Act: This law protects original works of authorship but does not protect a person’s identity itself. Therefore, a digital replica of someone’s voice or image does not count as copyright infringement—unless it copies actual content (like audio or visuals) that is already protected by copyright.
b. Federal Trade Commission Act (FTC Act): This law prohibits unfair or deceptive business practices. The Federal Communications Commission (FCC) has stated that if a digital replica misleads consumers, takes unfair advantage of a creator’s reputation, reduces the value of their work, or causes other serious harm, it could be considered deceptive or unfair under this law.
c. Lanham Act (Trademark Law): This may apply if a digital replica is used without permission in a way that confuses consumers—such as falsely suggesting that someone has endorsed a product or is associated with a company.
d. Communications Act: The FCC uses this law to regulate AI-generated voices in scam or harassing phone calls and allows state attorneys general to help enforce these rules.
The U.S. Copyright Office believes that, while existing federal laws offer some protection, they are clearly insufficient to address the broad range of harms caused by AI-generated digital replicas. For example, the Copyright Act does not protect a person’s identity; the FTC Act and Lanham Act mainly cover commercial deception or confusion, leaving non-commercial but harmful uses—such as deepfake pornography—largely unprotected.
In addition, trademark law protections usually apply only to well-known individuals, as it requires proof that consumers are likely to be confused about a person’s connection to a commercial product or service. The Communications Act is mostly limited to phone-based scams and does not cover wider online uses, such as user-generated deepfakes on websites and social media platforms.
2. Federal Legislative Proposals
In recent years, the U.S. Congress has paid increasing attention to the issue of digital replicas and proposed several legislative measures, such as the No AI FRAUD Act (No Artificial Intelligence Fake Replicas And Unauthorized Duplications Act) and the NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act). On January 13, 2025, Representative Rick W. Allen introduced the CLEAR Voices Act (Creating Legal and Ethical AI Recordings), which aims to establish technical and procedural standards for AI-generated voice replicas 5 .
In this context, the U.S. Copyright Office recommends creating a new statutory right specifically for digital replicas, with the following key features:
(1) Regulated Subject Matter: The law should target digital replicas that are so realistic they are difficult to distinguish from real individuals.
(2) Protected Persons: The law should protect all individuals, not just celebrities or those with commercial value.
(3) Infringing Acts: Infringement should focus on acts of public dissemination, such as distribution, public performance, broadcast, display, or making the replica publicly accessible.
(4) Exceptions: Personal use or use in the creative process (e.g., drafts or experiments) should not be considered infringement, in order to avoid overregulating artistic innovation.
Implications for Taiwan
Taiwan faces similar challenges, as there is currently no dedicated legal framework specifically addressing digital replicas. At present, regulatory efforts rely on existing legal provisions. For instance, creating or distributing AI-generated videos that synthesize a person’s face or voice without consent may infringe upon portrait rights or other personality interests protected under civil law. Collecting, processing, or using an individual’s facial features or voice without their consent may violate the Personal Data Protection Act. Where deepfake technologies involve the alteration of existing audiovisual content, such acts may be considered reproduction or adaptation under the Copyright Act. Additionally, the unauthorized use of a person’s image in advertisements could mislead consumers into believing the individual endorsed the product, potentially violating Article 25 of the Fair Trade Act, which prohibits deceptive or unfair conduct likely to disrupt trading order—such use may also constitute false advertising. Furthermore, if deepfake content imitates a registered trademark, it may amount to trademark infringement.
Given the above, Taiwan should assess whether its current legal framework provides sufficient protection. If gaps are identified, appropriate legal revisions should be considered. The formulation of such regulations will require broad social consensus and a careful balancing of technological progress and the protection of individual rights.
For Taiwan, the U.S. Copyright Office’s report serves as a valuable reference, underscoring the urgent need to establish a comprehensive regulatory framework for digital replicas in the AI era. In formulating relevant policies, Taiwan should:
1. Conduct an in-depth analysis of the protection gaps within the current legal system;
2. Solicit input from a broad range of stakeholders to build social consensus;
3. Strike a balance between technological innovation and the protection of individual rights;
4. Closely monitor international developments and adjust domestic policy directions accordingly.
[1] U.S. Copyright Office (July 2024), Copyright and Artificial Intelligence, Part 1: Digital Replicas (A Report of the Register of Copyrights). https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf (Last access date: June 17, 2025)
[2] U.S. Copyright Office (January 2025), Copyright and Artificial Intelligence, Part 2: Copyrightability (A Report of the Register of Copyrights). https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf (Last access date: June 17, 2025)
[3] U.S. Copyright Office (May 2025), Copyright and Artificial Intelligence, Part 3: Generative AI and Copyright Training Uses. [Pre-publication version] https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf (Last access date: June 17, 2025)
[4] Hern, A. (May 27, 2024), Scarlett Johansson’s OpenAI clash is just the start of legal wrangles over artificial intelligence, The Guardian. https://www.theguardian.com/technology/article/2024/may/27/scarlett-johansson-openai-legal-artificial-intelligence-chatgpt (Last access date: June 17, 2025)
[5] Allen, R. W. (May 22, 2024), Congressman Allen Re-Introduces Bill to Curb AI-Generated Robocalls, Office of Representative Rick W. Allen. https://allen.house.gov/news/documentsingle.aspx?DocumentID=6601 (Last access date: June 17, 2025)
1. Part1: Digital Replicas (July 2024) 1
2. Part2: Copyrightability (January 2025) 2
3. Part3: Generative AI Training (May 2025) 3
These reports aim to guide Congress, courts, and executive agencies toward a balanced approach—advancing technology while protecting human creativity. This summary highlights the first report on Digital Replicas and its implications for Taiwan.
Definition and Technological Context
“Digital replicas” refer to highly realistic reproductions of a person’s voice, likeness, or image using digital tools, often referred to as Deepfakes. With rapid AI advances, such content can now be created swiftly, convincingly, and without consent—raising new challenges for copyright, personality rights, and social trust.
Case Spotlight: Scarlett Johansson vs. OpenAI
In May 2024, OpenAI launched a real-time voice interaction feature for ChatGPT. One of the developed voices, “Sky,” drew public attention for its striking resemblance to the voice performance of renowned actress Scarlett Johansson in the film “Her.” Johansson expressed shock and anger and, through her attorney, formally sent a letter to OpenAI demanding: “Take down my voice immediately!” This incident highlights the potential personality rights issues that may arise from the commercial use of digital replica technologies. 4
Social Impact and Legal Gaps
Beyond the frequent targeting of celebrities and political figures, what alarms the public even more is the cloning of ordinary individuals' voices or images to deceive their friends and family in scams. This means anyone could become a victim of unauthorized digital replicas, posing a serious threat to the foundation of social trust and highlighting the urgent need for updates to the current legal framework.
Analysis of U.S. Legal Framework
1. The U.S. Copyright Office observed that public commentary overwhelmingly reflects a shared concern: existing U.S. laws are ill-equipped to govern the rising tide of unauthorized digital replicas. There is a pressing need for new federal legislation to meet the moment and respond to this rapidly evolving challenge.
The inadequacies of the existing U.S. legal framework and its regulations include:
(1) State Law
In the United States, most state laws regulate unauthorized digital replicas primarily through the Right of Privacy and the Right of Publicity. The Right of Privacy is intended to protect individuals from unreasonable intrusions, including false representations or the misappropriation of personal identity. Relevant torts under this right include Appropriation and False Light. The Right of Publicity, on the other hand, focuses on the use of personal identity elements—such as name, image, or voice—for commercial purposes, in order to safeguard their economic value. In recent years, some states have begun to enact or amend laws in response to this issue. For instance, Tennessee has extended the right of publicity to include protection for voice imitation, expanding its coverage even to non-commercial dissemination.
However, state-level regulations remain notably inadequate in addressing unauthorized digital replicas. The main challenge lies in the lack of consistency across states: laws differ in terms of who is protected (e.g., only celebrities or all individuals), whether protection applies only to commercial uses, and whether rights continue after death, among other variables.
(2) Federal Law
Although there is currently no specific federal law in the United States that directly regulates digital replicas, several existing laws may still apply.
a. Copyright Act: This law protects original works of authorship but does not protect a person’s identity itself. Therefore, a digital replica of someone’s voice or image does not count as copyright infringement—unless it copies actual content (like audio or visuals) that is already protected by copyright.
b. Federal Trade Commission Act (FTC Act): This law prohibits unfair or deceptive business practices. The Federal Communications Commission (FCC) has stated that if a digital replica misleads consumers, takes unfair advantage of a creator’s reputation, reduces the value of their work, or causes other serious harm, it could be considered deceptive or unfair under this law.
c. Lanham Act (Trademark Law): This may apply if a digital replica is used without permission in a way that confuses consumers—such as falsely suggesting that someone has endorsed a product or is associated with a company.
d. Communications Act: The FCC uses this law to regulate AI-generated voices in scam or harassing phone calls and allows state attorneys general to help enforce these rules.
The U.S. Copyright Office believes that, while existing federal laws offer some protection, they are clearly insufficient to address the broad range of harms caused by AI-generated digital replicas. For example, the Copyright Act does not protect a person’s identity; the FTC Act and Lanham Act mainly cover commercial deception or confusion, leaving non-commercial but harmful uses—such as deepfake pornography—largely unprotected.
In addition, trademark law protections usually apply only to well-known individuals, as it requires proof that consumers are likely to be confused about a person’s connection to a commercial product or service. The Communications Act is mostly limited to phone-based scams and does not cover wider online uses, such as user-generated deepfakes on websites and social media platforms.
2. Federal Legislative Proposals
In recent years, the U.S. Congress has paid increasing attention to the issue of digital replicas and proposed several legislative measures, such as the No AI FRAUD Act (No Artificial Intelligence Fake Replicas And Unauthorized Duplications Act) and the NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act). On January 13, 2025, Representative Rick W. Allen introduced the CLEAR Voices Act (Creating Legal and Ethical AI Recordings), which aims to establish technical and procedural standards for AI-generated voice replicas 5 .
In this context, the U.S. Copyright Office recommends creating a new statutory right specifically for digital replicas, with the following key features:
(1) Regulated Subject Matter: The law should target digital replicas that are so realistic they are difficult to distinguish from real individuals.
(2) Protected Persons: The law should protect all individuals, not just celebrities or those with commercial value.
(3) Infringing Acts: Infringement should focus on acts of public dissemination, such as distribution, public performance, broadcast, display, or making the replica publicly accessible.
(4) Exceptions: Personal use or use in the creative process (e.g., drafts or experiments) should not be considered infringement, in order to avoid overregulating artistic innovation.
Implications for Taiwan
Taiwan faces similar challenges, as there is currently no dedicated legal framework specifically addressing digital replicas. At present, regulatory efforts rely on existing legal provisions. For instance, creating or distributing AI-generated videos that synthesize a person’s face or voice without consent may infringe upon portrait rights or other personality interests protected under civil law. Collecting, processing, or using an individual’s facial features or voice without their consent may violate the Personal Data Protection Act. Where deepfake technologies involve the alteration of existing audiovisual content, such acts may be considered reproduction or adaptation under the Copyright Act. Additionally, the unauthorized use of a person’s image in advertisements could mislead consumers into believing the individual endorsed the product, potentially violating Article 25 of the Fair Trade Act, which prohibits deceptive or unfair conduct likely to disrupt trading order—such use may also constitute false advertising. Furthermore, if deepfake content imitates a registered trademark, it may amount to trademark infringement.
Given the above, Taiwan should assess whether its current legal framework provides sufficient protection. If gaps are identified, appropriate legal revisions should be considered. The formulation of such regulations will require broad social consensus and a careful balancing of technological progress and the protection of individual rights.
For Taiwan, the U.S. Copyright Office’s report serves as a valuable reference, underscoring the urgent need to establish a comprehensive regulatory framework for digital replicas in the AI era. In formulating relevant policies, Taiwan should:
1. Conduct an in-depth analysis of the protection gaps within the current legal system;
2. Solicit input from a broad range of stakeholders to build social consensus;
3. Strike a balance between technological innovation and the protection of individual rights;
4. Closely monitor international developments and adjust domestic policy directions accordingly.
[1] U.S. Copyright Office (July 2024), Copyright and Artificial Intelligence, Part 1: Digital Replicas (A Report of the Register of Copyrights). https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf (Last access date: June 17, 2025)
[2] U.S. Copyright Office (January 2025), Copyright and Artificial Intelligence, Part 2: Copyrightability (A Report of the Register of Copyrights). https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf (Last access date: June 17, 2025)
[3] U.S. Copyright Office (May 2025), Copyright and Artificial Intelligence, Part 3: Generative AI and Copyright Training Uses. [Pre-publication version] https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf (Last access date: June 17, 2025)
[4] Hern, A. (May 27, 2024), Scarlett Johansson’s OpenAI clash is just the start of legal wrangles over artificial intelligence, The Guardian. https://www.theguardian.com/technology/article/2024/may/27/scarlett-johansson-openai-legal-artificial-intelligence-chatgpt (Last access date: June 17, 2025)
[5] Allen, R. W. (May 22, 2024), Congressman Allen Re-Introduces Bill to Curb AI-Generated Robocalls, Office of Representative Rick W. Allen. https://allen.house.gov/news/documentsingle.aspx?DocumentID=6601 (Last access date: June 17, 2025)
