Generative AI in Music: Copyright, Fair Use and Ethics
Generative AI is a category of artificial intelligence that creates new content - the likes of images, text, or even music-based on user prompts. These models are trained on vast datasets and then produce outputs that closely resemble human creations.
Today, many generative AI companies train their models to replicate music to generate human-like music. We explored key legal concerns, including whether AI-generated music qualifies for copyright protection, how fair use applies to the training of AI models using copyrighted material, and the ethical implications of integrating AI into creative fields. Through interviews with Jessica Richard of the Recording Industry Association of America (RIAA) and Dr. Moiya McTier of the Human Artistry Campaign, we discuss these topics in depth, highlighting the challenges and opportunities posed by this emerging technology.
Copyright Law and RIAA’s Efforts
Under current U.S. copyright law, AI generated music that has been entirely produced using an AI model is not eligible for copyright protection. The U.S. Copyright Office provides detailed and valuable guidance on the copyrightability of works involving AI-generated content. According to their framework, copyright protection is reserved exclusively for creations that display human Authorship (U.S. Copyright Office). For a work containing AI-generated material to qualify, a human must make meaningful and substantial creative contributions that reflect originality. Actions like selecting prompts or making small edits to AI outputs are not regarded as sufficient. Instead, the human input must significantly transform the AI-generated material into a unique and original work that demonstrates creative intent (U.S. Copyright Office).
This clarification is essential in distinguishing passive use of AI from active authorship. What matters is the evidence of creative involvement that goes beyond the AI's automated process, shaping the final output in a way that reflects human originality. This approach aligns with copyright principles, ensuring that the protections offered by copyright law are reserved for those only who actively contribute to the final work (U.S. Copyright Office).
The Recording Industry Association of America (RIAA) has been proactive in addressing the legal challenges posed by AI companies, their resulting models, and musical products made through those models. RIAA represents major record labels and artists in the U.S., working to protect the intellectual property that powers the music industry. In the face of challenges brought by AI, the RIAA plays an important role in tackling copyright issues and fighting for policies that ensure artists’ work is respected and valued.
In June 2024, the RIAA, along with major record labels such as Universal Music Group, Sony Music Entertainment, and Warner Music Group, filed lawsuits against AI startups Suno and Udio. These companies were accused of using copyrighted music without authorization to train their AI models, resulting in outputs that closely mimic, or are ‘substantially similar’ to the voices and styles and copyrighted music of renowned artists like Michael Jackson and Bruce Springsteen. (Pitchfork)
Jessica Richard is an attorney who serves as the Senior Vice President of Federal Public Policy at the RIAA. She represents the RIAA on Capitol Hill and state legislative bodies, and before federal agencies, focusing on intellectual property and technology policy. Prior to joining the RIAA, she practiced law for nearly eight years at Williams & Connolly, LLP, specializing in commercial litigation and intellectual property. We asked her about the ownership of AI-generated music:
Is AI-Generated Music A ‘Derivative Work’?
A ‘derivative work’ is a creation that is based on or derived from one or more existing works, such as a musical arrangement. According to U.S. copyright law, a derivative work must incorporate substantial original authorship that transforms that existing material into a new, original piece. (Cornell)
If an AI system produces a song that mimics or is based upon a copyrighted artist's work, is it considered a derivative work? If so, it would still require authorization (a license) from the original copyright holder, as the new creation is built upon their copyrighted material. Without proper licensing, any AI-generated content that incorporates elements of existing songs risks infringing on the exclusive rights of the copyright holder. For example, did the AI-generated song mimicking artists Drake and The Weeknd violate those artists’ rights as to derivative works (Harvard Law School)? Because AI is a new technology and the U.S. courts have not yet ruled on many AI-related issues relating to music, many issues are unclear.
Jessica Richard shared her view on the question of derivative works:
Based on Richard's analysis, the core legal framework for AI in music centers on two critical points: first, examining whether AI companies have obtained proper permission to use copyrighted material for training their models (the "way in"), and second, scrutinizing how the AI outputs potentially violate both copyright and personality rights like voice and likeness (the "way out"). While derivative work arguments exist, Richard emphasizes that the most compelling current legal battles focus mainly on these input and output violations, suggesting this will be the key issue as courts prepare to shape AI laws.
Fair Use & Preservation Of Human Art
Fair use allows copyrighted material to be used without permission under certain circumstances. Fair use rests on factors including: purpose of use, nature of copyrighted work, amount of material used, and the market impact (Copyright Alliance). AI companies argue that using copyrighted music to train their models is fair use, particularly emphasizing that feeding copyrighted music into their models to generate new outputs is transformative.
The degree of transformation and the impact on the original work’s market value are critical factors in fair use analysis (ABA Journal). For example, AI-generated music that relies heavily on copyrighted inputs may not sufficiently alter the original work to meet the standards of transformation required under fair use. Additionally, when AI-generated material floods the marketplace, it risks undermining the commercial value of original songs — which, again, is a key consideration in the fair use doctrine (Harvard Westlake Chronicle).
Ms. Richard disagrees that AI companies engage in fair use:
Ms. Richard also noted that AI training involves using entire copyrighted songs, which could be flagged as excessive use under the ‘amount used’ factor, and under the factor of ‘market impact’ if AI-generated material floods the marketplace, this could undermine real artists and their art.
Dr. Moiya McTier is a scientist, storyteller, and communicator who brings a unique perspective to the conversation about AI and creativity. She studied both astronomy and mythology at Harvard University, becoming the first person to combine these fields, and went on to earn a PhD in astrophysics from Columbia University. She currently serves as Senior Advisor for the Human Artistry Campaign, a multi-sector global alliance of over 150 organizations, including organizations representing musicians, writers, songwriters, filmmakers, artists, and athletes, that advocates for responsible AI use and regulation. For example, the Campaign supports legislative measures like the NO FAKES Act, which aims to safeguard individuals against unauthorized digital replicas created by AI. The Campaign’s mission is essential: to ensure AI supports human artistry rather than replacing it. By promoting ethical AI practices, they aim to protect the livelihoods and rights of artists everywhere.
Dr. McTier expressed concerns about AI systems using copyrighted material without consent, emphasizing the necessity for transparency and proper licensing to protect artists' rights. Dr. McTier also stressed the negative impact AI can have in general and on real art if it manages to flood the market:
The following video by Steve Onotera (@samuraiguitarist) highlights how AI-generated music has advanced but still lacks the creativity driven by human emotional expression. He emphasizes the need to protect human artistry and ensure fair recognition for artists whose work supports AI training.
Licensing, Crediting And Ethics
Without proper licensing, artists and rights holders do not get the credit or compensation they deserve. Clear laws to ensure AI developers get explicit permission before using someone’s work would be best practice to ensure basic ethical standards and help protect the rights of creators. Both our interviewees had some beautiful stories to share regarding ethical, consensual use of AI in music relating to Randy Travis and the Beatles:
References
ABA Journal. "AI-Generated Music Is Everywhere. Is Any of It Legal?" ABA Journal, 2023.
https://www.abajournal.com/web/article/ai-generated-music-is-everywhere-is-any-of-it-legal
Copyright Alliance. “What is Fair Use?”
https://copyrightalliance.org/faqs/what-is-fair-use/
Harvard Law School. "AI Created a Song Mimicking the Work of Drake and The Weeknd. What Does That Mean for Copyright Law?" Harvard Law Today, 2023.
https://hls.harvard.edu/today/ai-created-a-song-mimicking-the-work-of-drake-and-the-weeknd-what-does-that-mean-for-copyright-law/#:~:text=In%20general%2C%20music%20in%20the,those%20outputs%20themselves%20are%20protected.
Harvard-Westlake Chronicle. "The AI Music Issue." Harvard-Westlake Chronicle, 2023.
https://hwchronicle.com/108719/opinion/the-ai-music-issue/#:~:text=AI%2D%20generated%20music%20is%20able,input%20of%20the%20copyrighted%20song.
Images by Jonathan Kemper, Growtika, 愚木混株 cdd20, on Unsplash.
Legal Information Institute. "17 U.S. Code § 101 - Definitions." Cornell Law School.
https://www.law.cornell.edu/uscode/text/17/101.
Madison Bloom. "Music Industry Groups Sue AI Companies for Stealing Artists’ Work to Generate Music." Pitchfork, June 2024.
https://pitchfork.com/news/music-industry-groups-sue-ai-companies-for-stealing-artists-work-to-generate-music.
SamuraiGuitarist. (2024). A.I. music is REALLY GOOD, and it's not a big deal [Video]. YouTube.
https://www.youtube.com/watch?v=aQC0FI_asKY.
U.S. Copyright Office. "Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence." Federal Register, March 16, 2023.
https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence.