Allegro

BIG MUSIC AND A.I.

How does it actually work?

Volume 124, No. 10November, 2024

William Meade

This month’s submission for Local 802’s A.I. series is by William Meade, the chair of Local 802’s A.I. Committee.

In last month’s article, I explored the steps that TikTok uses to gather vast amounts of music data to train and develop A.I. This month, I want to break down the relationship between the major music labels and big tech companies. And the biggest of them all is Universal Music Group (UMG), which houses some of the most influential and successful artists globally. It controls the rights to over three million recorded songs and the publishing rights for over four million compositions.

UMG, like other music companies, is navigating the legal, financial and ethical implications of using music for training A.I. models. Let’s break down the steps through which UMG makes their music available for A,I, training, the current practices regarding artist compensation and rights, and whether artists have any say in stopping or controlling this use.

UMG holds the rights to a vast catalog of music, which it often licenses out to various companies for different purposes, including digital streaming, advertisements, films, and — more recently — A.I. applications. One way UMG makes their music available for A.I. training is through specific licensing deals with technology companies. These agreements allow companies to use UMG’s music to train A.I. models for purposes such as:

Generating new songs
Improving recommendation algorithms on platforms like Spotify or YouTube
Enhancing audio analysis tools that can automatically detect musical qualities like mood, genre, or rhythm

The terms of these licensing agreements typically include details about how the music can be used, who has access to it, and whether the company will pay royalties or fees to UMG.

Data Sharing for Research and Development

In some cases, UMG may also collaborate with tech companies and A.I. developers in research and development partnerships. These arrangements are usually centered around A.I. innovation rather than direct profit. For instance, a tech company might partner with UMG to analyze musical trends or create tools that enhance user experience on music platforms. While this kind of data-sharing often benefits both parties, it involves using vast amounts of UMG’s catalog as training data for machine learning models.

One notable example is the partnership between UMG and Endel, an A.I. company that generates soundscapes for focus, relaxation and sleep. UMG artists like Grimes have licensed their voices and sounds to Endel for A.I.-driven music creation.

Streaming Platforms and A.I. Integration

UMG works with major streaming services such as Spotify, Apple Music and YouTube, which also rely heavily on A.I. for their recommendation engines. While these platforms do not necessarily “train” A.I. in the traditional sense, their use of A.I. to curate playlists and suggest new artists is built upon data derived from UMG’s catalog. Every time a user listens to a song, that data is fed into A.I. algorithms that are constantly learning and improving to offer better recommendations.

In these instances, the use of UMG’s catalog is often governed by existing digital licensing agreements, but it illustrates how A.I.-driven processes rely on music libraries to function.

Content ID and Copyright Management Tools

UMG, like other major labels, employs sophisticated A.I. tools to manage copyright claims across platforms like YouTube. The Content ID system on YouTube is an A.I.-based tool that scans videos for copyrighted material, ensuring that rights holders are compensated if their content is used. UMG allows its music catalog to be scanned by these A.I. systems to track and monetize instances where songs appear in user-generated content. While this practice is primarily aimed at copyright enforcement, it also shows how UMG’s music is integrated into A.I. systems to enhance their capabilities and manage intellectual property.

Can Artists Stop AI Training or Get Compensated?

Now the question many artists are asking is: Do they have any control over whether their music is used to train A.I. models, and can they be compensated for it? The answer lies in a few different areas, including contracts, legal protections and the evolving nature of copyright law.

Artist Contracts

In many cases, artists signed to major labels like UMG have already agreed to give the label control over how their music is used, including for licensing purposes. Depending on the contract’s specifics, UMG might have the legal right to license an artist’s work to a third party for A.I. training, even without the artist’s direct consent. However, if the contract does not specify the use of music for A.I. training, artists might have room to negotiate or push back. Some artists may try to revise their contracts to include explicit terms regarding the use of their music in A.I. applications, demanding compensation for such use or the ability to approve or deny these activities.

Royalties and Compensation

For artists who are aware that their music is being used in A.I. training, the next question is how they can be compensated. Typically, when music is licensed to a tech company for use in A.I. models, UMG receives a fee or royalties. Artists, depending on their contract, may receive a portion of these royalties. In some cases, artists may be entitled to mechanical royalties (which are generated from the reproduction of music) or performance royalties (which are generated when music is publicly played). However, the legal landscape for A.I.-generated content is still in its infancy, and determining what kind of royalties apply to AI training data remains an evolving conversation.

Laws and Regulatory Protections

With the rapid rise of A.I., regulatory frameworks are beginning to catch up. In the U.S., copyright law protects an artist’s original work, but the boundaries of that protection are being tested in the A.I. age. Currently, there are few laws that explicitly address the use of copyrighted material for training AI models. However, movements within the music industry are pushing for stronger regulations that ensure artists are fairly compensated when their work is used in this way. For example, the Recording Academy and other advocacy groups are lobbying for clearer guidelines on how A.I.-generated music should be handled under copyright law. This includes ensuring that artists receive compensation when their music is used to train A.I. models, as well as establishing rules around the creation and distribution of AI-generated content that mimics an artist’s style or voice. See Local 802’s A.I. Resource Page for a list of current pro-worker A.I. legislation.

The Role of Collective Bargaining

Artists may also have some protection through collective bargaining efforts. The AFM and the Recording Industry Association of America are calling for labels and tech companies to be transparent about how they use music in A.I. training and to ensure that artists are fairly compensated. The AFM included A.I. provisions in its latest Basic Theatrical Motion Picture and Basic Television Motion Picture Agreements: “The contract incorporates new protections to safeguard against the encroachment of generative artificial intelligence (A.I.), providing compensation for musicians when their performances are used in conjunction with A.I. tools, and ensuring human creativity remains at the heart of the industry.”

Artist Control and Alternatives

Independent artists who control their own music rights may have more flexibility in deciding whether or not their music is used in A.I. training. If an independent artist does not want their work to be used for A.I. purposes, they can simply choose not to license it to companies that engage in these activities. However, for artists under contract with major labels like UMG, opting out can be more challenging, as they often do not have direct control over how their music is licensed once they’ve signed over the rights.

A.I. and Music Industry Negotiations

The negotiation between music labels and A.I. companies is ongoing, and as the technology evolves, so will the legal and business frameworks surrounding it. Artists may begin to demand more control over how their music is used in A.I. training, potentially leading to new contract stipulations or industry-wide agreements that offer more protection.

Open Question

One last note. Many artists may come to find out that their music was “scraped” or “stolen” over the past 20 years as A.I. companies trained their models on any music they could find on the internet, included copyrighted or licensed works. What recourse do artists have? That is an open question…and one with high stakes.

Conclusion

Universal Music Group, like other major music labels, is navigating the complex landscape of A.I. and music. By licensing music for A.I. training and using A.I .for copyright management, UMG is both capitalizing on the technology and pushing the boundaries of how music is used in the digital age. For artists, the key issues revolve around control and compensation. While many artists currently lack the ability to stop their music from being used in A.I. applications, evolving contract negotiations, collective bargaining, and changes in copyright law could provide more avenues for artists to protect their work and receive fair compensation in the future.

We need everyone involved to make this happen.

Multi-instrumentalist William Meade, a member of Local 802 since 1979, is the chair of Local 802’s A.I. Committee. William is a seasoned performer and producer with over three decades of experience in television, Broadway, concerts and global events. During the last 12 months, he has spent considerable time and effort on understanding the accelerating impact AI is having across all professions, including that of musician’s. His knowledge of AI encompasses practical, academic and regulatory aspects, all of which continue to evolve at a rapid pace. E-mail William Meade at: wmeade@stellationentertainment.com. Send feedback on Local 802’s A.I. series to Allegro@Local802afm.org.

CLICK HERE TO JOIN OUR A.I. COMMITTEE E-MAIL LIST AND TO LEARN ABOUT UPCOMING MEETINGS