This blog post is adapted from an article written by Copyright Office staff that was originally published in WIPO Magazine’s special edition on Music and IP, April 2025, a publication of the World Intellectual Property Organization (WIPO).
The capabilities of the latest generative AI technologies raise significant questions about the nature and scope of human authorship—including in music. In early 2023, the U.S. Copyright Office launched an initiative exploring the intersection of copyright and artificial intelligence. In August 2023, the Office formally sought public input on the full range of copyright issues that have been raised by generative AI, receiving more than 10,000 comments from all 50 states and from 67 countries representing a broad range of perspectives. After reviewing these comments, the Office began working on a Report for Congress and the public, which can be found online at copyright.gov/ai.
To discuss the Office’s findings and latest decisions, Miriam Lord, Associate Register of Copyrights and Director of Public Information and Education, sat down with two colleagues, Senior Counsel Chris Weston and Assistant General Counsel Jalyce Mangum.
Miriam Lord (ML): Chris, what makes the U.S. copyright system unique when it comes to AI?
Chris Weston (CW): Original works of expression are protected under national laws, with protection in a particular country dependent on that country’s laws. International copyright conventions and treaties, like those through WIPO, establish obligations that provide more certainty about protection levels across countries.
When we talk about music, it’s important to understand that even in a single song, with rights in both the musical composition and sound recording, these rights are often divided among multiple rightsholders.
U.S. copyright law differs from copyright laws of other countries in several ways, including that it creates a system where applications to register copyrights are examined at the Office. Registration is not mandatory, but it does provide significant benefits. When it comes to copyrightability issues, including works containing AI-generated material, the Office is a kind of “natural laboratory” because our examiners deal with these issues daily.
ML: Jalyce, tell us about the Office’s study on AI and copyright law.
Jalyce Mangum (JM): The Office launched a broad AI initiative to address these emerging questions and issued guidance to help authors understand how to apply for copyright registration when their works included AI-generated material. We hosted public listening sessions and webinars, met with experts and stakeholders, and published a notice of inquiry seeking input from the public.
After reviewing more than 10,000 comments—an astounding response that helped inform our conclusions—we began working on a Report for Congress and the public. We published Copyright and Artificial Intelligence, Part 1: Digital Replicas, on July 31, 2024. Part 2: Copyrightability, published January 29, 2025, focuses on the copyrightability of outputs created using generative AI. The Office will issue subsequent Parts on the legal implications of training AI models on copyright-protected works, including licensing considerations and the allocation of potential liability.
ML: What were some of the common concerns the music community raised during the Office’s consultations?
JM: The feedback from performers, songwriters, composers, publishers, record labels, trade organizations, and advocacy groups was crucial in designing a comprehensive approach for our recommendations. Their concerns focused on five themes but also concurred with the broad consensus about safeguarding creatives and their works.
Specifically, their comments centered first on human authorship and AI’s impact on human creativity. Second, labor effects—they fear that rapidly generated AI content could overwhelm human-authored works in the marketplace. Third, they want control and compensation for the use of both their personas and copyright-protected works. Fourth, they recognize innovation aspects—that AI can assist the creative process and even enable performances by deceased artists. Finally, regarding licensing, they consider what type should be preferred: voluntary, collective, or compulsory. In short, should they have to opt in or opt out?
ML: What are unauthorized digital replicas, and what approach does the Office recommend?
JM: In 2023, a song dropped that appeared to feature the voices of Drake and The Weeknd, drawing more than 15 million social media views and 600,000 listens on Spotify. This “Fake Drake” incident, as some called it, showed how rapidly generative AI technology has become sophisticated and accessible enough that minimal expertise is required to produce convincing digital replicas.
Throughout our study, we heard from members of the music industry that they worried they would lose income due to the surge of voice clones or that the use of AI in sound recordings could displace human labor.
In the United States, a patchwork of laws governs individuals’ likeness and voice rights, but there are gaps and inconsistencies in who is protected and against what types of conduct. We concluded there was an urgent need for a new federal right protecting all people—not just celebrities—from the unauthorized use of their likeness and voice.
ML: The Office recently registered a sound recording by Grammy-winning artist Randy Travis. Using his AI-assisted recording as an example, how do you distinguish between AI as a creative tool versus a substitute for human creativity?
JM: Artists have used technology for decades to enhance, modify, and add to their creations—that isn’t new, nor is the requirement for human authorship to secure copyright protection. The Office has registered more than a thousand works where applicants have followed our guidance to disclose and disclaim AI-generated material. In the copyrightability analysis, distinguishing between using AI as a tool to assist in the creation of works and using AI to stand in for human creativity is important. The difference is whether AI is enhancing human expression or is the source of the expressive choices.
In Part 2 of our Report, the Office affirmed that copyright does not extend to purely AI-generated material or material where there is insufficient human control over the expressive elements.
Randy Travis’s recent sound recording, “Where That Came From,” featuring an AI clone of his voice, is a good example of AI assistive use. Due to health challenges, Travis has limited speech function, but to realize the dream of a new record, a production team took the vocal track of a human singer and used AI as a tool to modify the sounds to match Travis’s iconic voice.
ML: Let’s zoom back out to the international context. Chris, how are other countries approaching these copyright issues with AI?
CW: We’re aware we can’t confine our attention to within U.S. borders since AI systems are trained, developed, and deployed around the world. Other countries are also considering these issues, and while identical approaches are unlikely, some level of consistency will contribute to smooth global commerce.
We are seeing similarities and differences in how countries are approaching these issues. In the United States, copyright protection requires human authorship, and other countries follow that model as well.
In China, a court recently recognized copyright in a picture created using AI, finding a human author exercised sufficient creativity in prompting the AI tool and then revising the output. And South Korea has registered an AI-created film as a compilation, based on the human creativity that went into the selection, coordination, and arrangement of its AI-generated components.
We continue to discuss ideas and experiences with our global counterparts. Areas of consensus are emerging and so are potential differences.
ML: Thanks, you two. Certainly a lot to dive into.
But speaking of deep dives, the Copyright Office maintains a central resource for copyright creators and users in music: Engage Your Creativity: What Musicians Should Know About Copyright.
This resource provides an overview of copyright law and protections available in the music context and includes videos and handouts. It addresses the two types of copyright-protected works in a recorded song, the process for copyright registration, and considerations when using someone else’s work. We also cover what a copyright owner can to do if a work is used unlawfully and provide an overview how the Music Modernization Act helps songwriters get paid.
Copyright law protects musical works and sound recordings the moment the artist fixes their creative work in a tangible format—for example, when a musical composition is notated in sheet music or a song is recorded in an audio file. That protection comes with exclusive rights giving copyright owners control over the use of their works, including the rights to publicly distribute and perform them, with public performances of sound recordings limited to digital audio transmissions. Anyone wishing to use the work in these ways must have the copyright owner’s permission, a licensing contract, or rely on a limitation or exception provided in the Copyright Act. Registering a musical work or sound recording with the Copyright Office is not required, but doing so provides additional benefits.
To learn more, visit us at copyright.gov.