Artificial Intelligence, or AI, has been and continues to be one of the biggest topics of conversation in the music industry since the advent of digital music distribution. Those around during the early 2000s will never forget how every conversation about music was dominated by the great debate of whether Napster would ruin, or save, the music industry.
While Napster (www.napster.com) still exists, the Napster as it was, was phased out by legitimate digital distributors who recognized and properly licensed the various rights embodied in sound recordings, along came streaming, once again upending the ecosystem, leaving an industry scrambling to realize fair compensation and protection of creators’ and owners’ rights.
Now here we are twenty-five years later; once again the music industry isn’t fully prepared, nor definitively protected by current laws designed to protect creators and performers, for this technological innovation that promises meteoric leaps in creativity as well as incredible financial possibilities.
GENERATIVE AI
Generative AI is a subset of artificial intelligence that creates “new content;” everything from images, to videos, to music. The output of generative AI tools is based on patterns and structures the model learns from inputting existing data it gleans from “scraping” or “mining” the already existing data found on the Internet.
For the music industry, that input is existing lyrics, melodies, and recordings with the intent of having the output of original lyrics, original melodies, and/or original recordings.
Much like the dawn of digital music and streaming, many of us became aware of the potential impact of generative AI once the proverbial cat had been let out of the bag. That moment for many of us was the release of the viral track “Heart on My Sleeve.”
Join Our Community
Be the first to read new articles, industry news, and more. Sign up to our newsletter today!
HEART ON MY SLEEVE
In April 2023, Ghostwriter977 released “Heart on My Sleeve”, a track he made using generative AI to simulate the music of Drake and The Weeknd. Universal Music Group, which has the exclusive rights to Drake and The Weeknd recordings, quickly alleged a copyright violation to get the various streaming platforms such as Spotify and Apple Music to take “Heart on My Sleeve” down but the damage was already done: a shorter clip of the song had gone viral on TikTok.
The track itself showed us how advanced generative AI had become in mimicking the music of actual artists and arguably became generative AI’s first viral moment.
Since the release of “Heart on My Sleeve” many tracks using simulations of other artists’ voices have released: Kanye’s “voice” performing a Plain White Ts’ song, Freddy Mercury’s “voice” “singing” Thriller and the “voice” of Rhianna covering Beyoncé.
Generative AI is “learning” exponentially; tracks which sound like they are featuring the voices of real artists are being created at lightning speed.
Yet, with the promised creative and financial potential, comes a host of legal and ethical considerations.
LEGAL CONSIDERATIONS
Generative AI companies train their AI tools by scrubbing the internet for data which, in the music industry, means ingesting large numbers of preexisting recordings and lyrics to produce new recordings and lyrics. There are several issues with this practice, from the legal concepts known as copyright infringement to appropriation of personality.
Copyright in information used in training models
Inherent in copyright is a bundle of rights including, among other rights, the copyright owner’s right to determine how their work is used commercially and how it may be reproduced. It appears that AI companies are making reproductions of copyrighted music to train their models for commercial gain.
There is Canadian defense to copyright law violations called “fair dealing” and a US defense called “fair use”. While different, these legal doctrines essentially allow portions of copyrighted material to be reproduced in certain circumstances without the permission of the copyright owner. This is a protection allowed in limited circumstances where among other considerations, value to society outweighs the enforcement of individual copyright holders’ rights. Those circumstances have been found to include education, research, news reporting, criticism and parody.
Copyright in the output of generative AI models
Also inherent in the bundle of copyright rights is the protection against having your works copied. When training generative AI with an artist’s pre-existing songs and recordings and then prompting the AI to create a song or recording that sounds like the training songs / recordings, the potential is ripe for a claim of copyright infringement on the basis that the output of AI steals from the lyrics, melodies and recordings that the model was trained on.
While AI companies appear to be flaunting the rights of copyright owners (the intricacies of the training models have been kept relatively non-transparent), the question of whether the output of generative AI models can or should be offered copyright protection is an issue.
Appropriation of personality
This practice of training AI with an artist’s voice and performance to create a soundalike also involves the unauthorized appropriation of personality.
Personality rights in Canada, come from a common law tort of appropriation of personality. The provinces of British Columbia, Manitoba, Newfoundland and Saskatchewan have each taken additional steps and enacted legislation governing personality. The tort of appropriation of personality, and the legislation enacted relating to the concept, protect against the unauthorized use of an individual’s persona, which includes attributes such as a person’s physical appearance, or voice.
Additional considerations
Unfortunately copyright law and the prohibition against appropriation of personality cannot fully protect artists from the impact of AI as we learned from Canadian artist, Ian Janes.
Janes recently learned that a new album under his name had been added to his Spotify artist profile, but he had not released new music. Someone had released an AI generated album under his name and had somehow succeeded in posting the album to Janes’ Spotify profile. While he has been successful in having the album removed from his profile, the album is still being exploited on Spotify.
Unfortunately, he did not have a recourse under copyright law for what was generated because his recordings, his compositions or other copyrights were used in the album that was AI generated. Furthermore, while he may have a recourse under the law that prohibits appropriation of personality for use of his name in connection with the AI generated album, he would have to institute a civil suit against the album owners and, from recent jurisprudence, his damages would be limited to what the album owners should have paid him if they had properly licensed his name.
THE ECONOMIC IMPACT:
In December of 2024, the International Confederation of Societies of Authors and Composers (CISAC) released a new report: the “Study on the economic impact of Generative AI in Music, and Audiovisual industries.” In particular, it looked at the potential losses to creators both from unauthorized unremunerated use of their copyrighted materials and from the competition from Generative AI music.
The study further found that, according to current projections, generative AI music will account for an estimated twenty percent (20%) of revenues for music streaming platforms and approximately sixty percent (60%) of revenues for music libraries.
While creator’s are facing a loss of Ten Billion British Pounds (€10,000,000,000) over five years if the current regulatory system remains unchanged, the study estimates that Generative AI music will be worth a cumulative Forty Billion British Pounds (€40,000,000,000) in that same time period. In addition, while revenues of twenty four percent (24%) of creators’ will be threatened in 2028, the yearly value of Generative AI music will reach around Sixteen Billion British Pounds (€16,000,000,000)in that same year.
The study concludes the economic risk to creator’s can be prevented by Generative AI companies clearing, and fairly compensating for, the rights to the copyrights used by generative AI providers.
THE “3 Cs” OF ETHICAL AI – CONSENT, CREDIT, AND COMPENSATION
While the proverbial cat is out of the bag, some rightsholders are working through the court system and legislation to try to limit the fallout in creative industries. Ultimately a mixture of litigation, legislation and licensing will likely be what is necessary to protect the rights of the various stakeholders in the music industry.
At the core of this movement, rightsholders in Canada and the US agree that consent, credit, and compensation are necessary for ethical AI, namely that AI companies need to seek and receive the consent of copyright owners and artists, appropriate credit must be afforded, and compensation must be provided.
The approach to achieving the so-called three Cs in the US has been piecemeal: there are several high visibility cases currently working their way through the courts. Some rightsholders are entering into direct licensing agreements with AI companies, and one state, Tennessee, has gone even further, enacting the Elvis Act; the first enacted legislation in the US specifically drafted to protect musicians from the unauthorized use of their voices through AI (i.e. a protection against appropriation of personality).
The three major music labels in the US (which control the majority of the worldwide English speaking music market) filed lawsuits against AI music firms Suno and Udio in June, 2024. The majors alleged that the widespread infringement of copyrighted sound recordings at such a vast scale to train AI models that would generate content that would compete with and ultimately drown out the original music on which the services were built.
In Canada, the Federal Court of Canada is currently being asked to make the declaration that copyright protection only be afforded to human created works. In early December 2024, six Canadian media companies, including CBC/Radio-Canada, the Globe and Mail, and the Toronto Star, filed a joint lawsuit against OpenAI, the owner of ChatGPT, for copyright infringement, alleging that OpenAI has taken significant portions of the outlets’ content for its training data without regard for their copyright in the content or the terms of use relating to the use of the content.
In addition, the Coalition for the Diversity of Cultural Expression (CDCE), which represents over 360,000 creators and nearly 3,000 cultural businesses recently provided a response to the Canadian Government’s consultation paper on Artificial Intelligence (AI) and Copyright. Their submission provided for three main recommendations:
- There should be no weakening of copyright protection for works currently protected (i.e., no fair dealing exception for text and data mining to use copyrighted works to train AI systems without the copyright owner’s consent).
- Copyright must continue to protect only works created by humans. The British Government has taken this approach, and the US Copyright Office has resisted attempts to register fully AI-generated (as opposed to AI-assisted) works.
- AI developers must be required to be transparent and disclose what works have been ingested as part of the training process. Without transparency any future of legitimate licensing will be futile.
The CISAC study concludes that failure to address the current risks presented if the currently regulatory framework, the risks to long-term effects on creators’ revenues, cultural diversity, and the quality of new music available to the public will be significant.
ETHICAL AI COMPANIES
Not all AI companies are created equally. There is a new wave of companies allegedly seeking to do business ethically:
- Jen, an AI music model by Futerverse, is trained on forty licensed catalogues and 150,000,000 songs, and is in the alpha stage. In keeping with rightsholders’ request for transparency, it will generate a hash on blockchain to verify the source and the output. There may still be personality rights considerations if their licenses do not include such rights.
- Rightsify, an AI music startup that offers music datasets which were licensed through direct agreements with musicians, has developed a model to create AI-generated music for hotels.
Keep in mind, there is still the issue of whether the output of these models can or should be copyrighted in Canada.
FINAL CONSIDERATIONS
Even if the copyright issues and personality rights issues are resolved in favour of the rightsholders, one still must consider the economic impact mass generated content will have on creators. For example, how can individuals who have to pay for musicians, instruments, gear, producers and other costs in connection with the creation of recordings compete with something that is created with the push of a button?
When seeking to have “Heart on My Sleeve” removed from music distribution platforms, a spokesperson for Universal Music Group stated something that continues to resonate with me:
“The training of generative AI using our artists’ music (which represents both a breach of our agreements and a violation of copyright law) as well as the availability of infringing content created with generative AI on DSPs, begs the question as to which side of history all stakeholders in the music ecosystem want to be on: the side of artists, fans and human creative expression, or on the side of deep fakes, fraud and denying artists their due compensation.”
© 2025 Edwards Creative Law, LLP
Updated to February 5, 2025
Edwards Creative Law is Canada’s Entertainment Law Boutique™, providing legal services to Canadians, and international clients who partner with Canadians, in the Film & Television, Music, Video Games and Apps, Publishing and Literary industries.
For more information or to set up a minute Discovery Call with one of our entertainment lawyers please feel free to Contact Us.
* This blog is for general informational purposes only and is not to be construed as legal advice. Please contact Edwards Creative Law or another lawyer, if you wish to apply these concepts to your specific circumstances.
Check out our popular blog posts:
Neighbouring Rights in Canada – Being a Musician is a Business
Setting up a Music Publishing Company in Canada
Copyright Protection & Classical Music
Work Made for Hire Explained
10 Co-Production Considerations in Canada – Ask an Entertainment Lawyer
Film Profits & Points – Ask an Entertainment Lawyer
The “Just Trust Me” Legal Agreement
Learn more about our services:
Model and Talent Law
Film and Television Law
Music Law
Video Game and App Law
Publishing and Literary Law
Employment Law
Dispute Resolution and Litigation Law
Corporate Law
International Services
