The music industry has experienced a significant transformation in the digital age, transitioning from traditional physical media to digital formats and streaming platforms.¹ The digital age has empowered musicians to produce and distribute their work with greater ease and efficiency, while also giving rise to new business models, disrupting traditional revenue streams, and challenging conventional notions of ownership and access in the music industry.² The emergence of streaming platforms such as Spotify, Apple Music, and others has transformed how people access music, providing an extensive library of songs that can be enjoyed anytime and anywhere with an internet connection.³

Artificial Intelligence (AI) has revolutionized streaming by effortlessly curating music playlists and tailoring song selections to individual preferences. This is made possible through the AI-based music recommendation system (MRS).⁴ A recommendation system filters and predicts a user’s preference for a specific element such as a song. It forms the core of large engines powered by recommender algorithms, suggesting items based on these predictions.⁵ An example of a platform built around a song recommendation engine is TikTok, whose unique algorithms promise creators many opportunities to grow organically.⁶

Popular music recommendations are primarily content-based and collaborative.⁷ The content-based approach recommends songs by extracting keywords from the descriptions of songs a user likes, comparing them with other tracks, and suggesting those with similar features.⁸ On the other hand, the collaborative approach relies on users’ overlapping preferences and ratings, suggesting songs that users with similar tastes enjoy.⁹ This method enables more personalized recommendations and is generally considered more accurate due to its focus on user interactions rather than content similarity.¹⁰

The theme for World Intellectual Property Day 2025, IP and Music: Feel the Beat of IP, underscores the evolving relationship between music and technology. Considering the profound impact of digitization across creative industries, this blog explores the intellectual property implications of music recommender algorithms, particularly within streaming platforms. It also examines the ethical considerations that arise in the digital age, where algorithmic decision-making increasingly shapes how music is created, distributed, and consumed.

Exploring the IP Challenges of Algorithm-Driven Music Consumption

It is crucial to note that before making music recommendations to a user, platforms utilize machine learning algorithms to process extensive datasets that include music genres, artists, lyrics, and user interactions.¹¹ The use of copyrighted materials like song lyrics and music compositions without proper authorization for training AI models can lead to infringement claims. For instance, in 2024, Sony Music Group issued notices to over 700 tech firms and music streaming platforms, cautioning them against using its music to train AI models without explicit authorization.¹² In a statement issued by the company, it emphasized that songwriters and recording artists rights including copyright should be respected.¹³

As users of music platforms indulge in curated playlists tailored to their preferences, it is equally vital to examine the impact of music recommendation systems on the royalties earned by rights holders. In essence, royalties are the payments made to rights holders for the use of their music, encompassing creators such as songwriters, musicians, performers, and other key stakeholders in the music industry.¹⁴ A recording royalty is the most common type of royalty that an artist and their recording label receive when their recording is downloaded digitally or streamed on a streaming platform.¹⁵ Royalties are important as they provide creators with passive income, encourage investment, and allow ownership retention, all while offering growth potential through the scalability of IP assets.¹⁶

The positive impact of recommendation algorithms lies in their ability to promote musical discovery, bringing attention to songs that might otherwise go unnoticed, thereby contributing to a more equitable distribution of royalties.¹⁷ On the flipside, recommendation algorithms have faced criticism for prioritizing mainstream songs, often sidelining lesser-known artists.¹⁸ This results in a narrowing of musical tastes and, inevitably, an unequal distribution of the revenues generated by streaming.¹⁹

Further, there is a growing concern that AI-generated music could flood streaming platforms.²⁰ This technology harnesses artificial intelligence algorithms to replicate musical elements such as melody, tempo, pitch, and range, facilitating the production of entirely new tracks.²¹ In 2024, it was reported that some Spotify users received AI-generated music recommendations.²² Notably, an AI-generated song even appeared on a German pop chart and has already amassed over 4 million streams on Spotify.²³ Therefore, depending on a listener’s preferences and a comparative analysis of previously played songs, recommendation algorithms may suggest AI-generated music and include it in the playlist.

Legal Uncertainty Around AI-Generated Music: Infringement and Authorship

Even though AI-generated music may align with a listener’s preferences, it raises important intellectual property concerns. The first concern revolves around potential infringement, encompassing three primary aspects: training data, stored information, and the generation of AI output. Firstly, if training data is obtained from unlicensed sources, such as by web scraping without the consent of rights holders, its use may infringe upon copyright and database rights.²⁴ Additionally, where AI generates and stores copies of the infringing material, rather than simply learning the relational principles from the training data, this could also constitute a violation of the reproduction right protected under copyright and database rights.²⁵ When AI-generated music reproduces a substantial portion of a copyrighted work, the output is a clear infringement.²⁶ However, when AI-generated music is “inspired” by earlier works and merely applies the relational principles it learned from analyzing copyrighted material to create something “new,” the legal assessment of infringement becomes more complex and uncertain.²⁷

The issue of ownership of AI-generated content also plays a significant role especially if copyright subsists in the output. In Kenya, where a literary, dramatic, musical, artistic work or computer program is computer generated, the author will be the person by whom the arrangements necessary for the creation of the work were undertaken.²⁸ Whilst authorship of AI-generated content may be straightforward in theory, given that AI cannot own or claim authorship or copyright under Kenya’s current legal framework (as Section 2 of the Copyright Act, 2001 limits protection to natural and legal persons), it may not be the case in practice. In practice, identifying the author is complicated due to the involvement of multiple parties, such as the software’s creator, those who trained the system, and the users who prompted it to generate the material.²⁹

Music Recommendation Systems and Publicity Rights

Moreover, music recommendation systems may also suggest music that mimics the voices of popular musicians. For instance, in 2023, a track titled “Heart on my Sleeve” that claimed to feature Drake and the Weeknd racked up millions of plays across TikTok, Spotify and YouTube before it was removed.³⁰ The track was not a collaboration between the two artists but rather an AI generated song by a Tiktok user who had trained the AI on their music styles.³¹ In cases where the voices of musicians are mimicked in AI soundalikes as illustrated above, they may turn to the right of publicity to protect the use of their voice.³² According to the International Trademark Association, the right of publicity is an intellectual property right that protects against the unauthorized commercial use of a person’s name, likeness, voice, or other identifiable attributes.³³ The voices of singers have previously been protected by courts especially in the case of Midler v Ford Motor Co³⁴ and Waits v. Frito-Lay Inc.³⁵ In both cases, the court ruled that the commercial use of vocal imitations of well-known singers amounted to misappropriation of their voices.³⁶

Ethical Implications of Personalised Music Discovery

To accurately predict users’ preferences, recommender systems rely on a large amount of user data that may be collected beyond users’ expectations.³⁷ The information shared by users may include ratings or comments, posting personal information on their profile, making purchases³⁸ or demographic information.³⁹ The use of this data raises privacy concerns about ethical data practices. For instance, regarding data collection, users often overlook how much data service providers collect and derive, mainly because privacy policies are rarely read and online activity is routine. In some instances, users may not manage to opt-out of such data gathering.⁴⁰ Additionally, online data is often difficult to remove, and providers may intentionally prevent deletion because user information holds commercial value for competitive analysis or sales.⁴¹ Beyond that, data that seems erased may still exist elsewhere in the system, such as in backups, resulting in retention beyond its intended lifespan.⁴² This practice raises critical concerns regarding the principles of purpose limitation, data minimization, and storage limitation.

Finally, a significant ethical concern is the risk of bias and the reinforcement of stereotypes or discrimination if recommendation systems are trained on data reflecting existing social inequalities, such as race or gender.⁴³ This emphasizes the need for fairness in algorithm design and implementation. Another issue is the vulnerability of these systems to be manipulated by malicious actors with hidden agendas, who may exploit the algorithms to push specific content or promote harmful ideologies.⁴⁴ This threatens the reliability of the recommendations users receive.⁴⁵ Additionally, the lack of transparency in recommendation systems adds complexity to the ethical concerns surrounding them.⁴⁶

Conclusion

The digital transformation of the music industry highlights critical intellectual property and ethical issues. AI-driven recommendation systems bring challenges such as unauthorized use of copyrighted materials, ambiguous ownership of AI-generated music, and unequal royalty distribution. Ethical concerns, including data privacy, algorithmic bias, and potential exploitation, further complicate the landscape. Moving forward, it is essential to refine existing frameworks to address the evolving challenges posed by emerging technologies, balancing the protection of both creators’ and users’ rights with ethical considerations. This will protect creators’ intellectual property, safeguard users’ access and privacy, and navigate the complexities of AI and data usage to promote a fair and accountable digital music ecosystem.

The image is AI generated via Microsoft Designer