Although music recognition apps such as Shazam have become embedded in our culture, AI has only become more advanced enhancing our ability to document, manipulate and market music into an unprecedented era.
Many AI music generators learn from countless copyrighted songs, without seeking consent or without the author receiving monetary compensation, giving the AI tools to create ‘art’ mimicking artists’ distinct styles. American AI advocates argue that data training falls under the “fair use” of copyright because the creations are transformative, do not create substantially similar works, and have no impact on the original work’s market.
Even still, can songs created by an AI be copyrighted? Can the owner of the copyrighted material used to train the AI claim the copyright of the work?
![](https://www.thomann.de/blog/wp-content/uploads/2023/01/AImusicheader-770x425-1.png)
Traditionally Intellectual Property is granted to the songwriter or artists and allows them to own the rights to the music they have contributed to. The US Copyright Act protects “works of authorships” allowing human authors to secure exclusive rights, thereby stating that animals and natural forces cannot hold copyright. The UK is one of few countries that have issued copyright for works generated by an AI, noting that the author is “the person by whom the arrangements necessary for the creation of the work is undertaken.”
This raises the question of who made these arrangements and who could be considered a contributor.
The contributors to the AI creation could include
- The creator of the AI codes and algorithm
- The copyright owners of data it is trained on
- The processors of said data
- The end-user
In 2007 The Yellow Pages faced an early copyright case involving algorithmically generated content in Australia. After discovering their competitor was lifting from their database, they sued for copyright infringement. The case was dismissed as the phone book was largely assembled by an algorithm with minimal human effort, rendering it ineligible for copyright protection. Therefore, AI music will likely be devoid of copyright protection in Australia, given the legal precedent.
AI creations could be considered derivative works in the US due to the copyrighted material needed for data training. Though, AI outputs may contain no part of the original data, likeness to copyrighted works could spur an infringement case. Lawsuits for the alleged likeness of copyrighted works have been successful before. The Marvin Gaye estate sued Robin Thicke and Pharrell Williams over their song Blurred Lines focusing on the ‘feel’ of the song in 2015 and Bette Midler sued Ford Motor Co. after they used another singer to imitate the likeness of Midler’s voice in an advertisement.
Michael Nash, EVP and chief digital officer at Universal Music Group offered solutions in an article for Music Business Worldwide. Policymakers need to continue ensuring legislation supports copyright protection for artists without hindering technological innovation. Potentially involving new licenses providing compensation for artists involved in AI training sets. The industry needs to agree on standards and approaches under existing legal frameworks while considering independent artists and businesses without the legal or financial resources to keep up.
This sector in the music industry is constantly evolving with discoveries emerging all the time. I will continue to follow this trend in my next blog posts, looking at new companies and how artists are interacting with AI.