Below, Meller sets out his principles for ethical AI music licensing — built on consent, fair compensation and transparency, with artists’ vocal rights off the table.
Ready or not, the AI revolution in music is here. Generative AI has been used in recordings for a while now in various disclosed and undisclosed manners. But the real story of the year in this space is the progress of licensing arrangements between rightsholders and AI companies—including Suno, Udio, Klay and Spotify’s AI platform — who seek legitimization for their offerings.
Generative AI use in music creation must put songwriters and recording artists first. Their life’s work is the backbone of our industry and, if licensed, the golden ticket to potential success for AI music offerings. This article will lay out some key considerations for an ethical framework to our rapidly developing AI landscape.
Preface: Understand the Moment
Among music creators, there is a lot of excitement — and skepticism — about AI tools. But in my conversations, many agree that it can be an effective complement to an existing skillset. It still needs human interaction and expertise. The doomsday scenario of the robots taking over is still a way off.
The greatest unforced error of the piracy era was the time it took for the industry to redirect its focus from misdirected punishment (who can forget the series of RIAA filesharing lawsuits against innocent grandmothers?) to embracing the Internet era and working towards a solution. We’re at an inflection point where greater AI use is a foregone conclusion. So let’s dig in.
Approvals are Non-Negotiable
The most spirited legal debate around generative AI at the moment surrounds whether record labels and music publishers have the right to license music rights to AI services without recording artist/songwriter approval. Nearly every publishing and recording agreement gives the company the right to issue blanket licenses while reserving certain approval rights for the grantor. Which one wins out?
It’s a trick question; regardless of which way a contract leans on a read of the language, recording artist and songwriter permissions are essential to any ethical, commercially viable AI offering. Generative AI would be nothing without the ability to train off the genius of creators. So, let’s start from a position of respect and consent.
Recent news of Universal Music Group and Spotify’s deal for an on-platform AI service reports that UMG writers and artists will have to affirmatively opt-in to participate. That’s heartening news and aligns with a creator-first perspective.
The real test will be how those companies and other players follow through on that mandate, and the details of the approval and compensation structure. Meanwhile, other labels and publishers have publicly announced partnerships with AI services while deciding not to seek artist/writer permission. Even worse, certain labels are insisting on language in new deals stating they have full rights in perpetuity to both artificially generated voice and name, image, and likeness reproductions. Yikes.
For music attorneys negotiating artist and songwriter deals going forward, the approvals clause has been upgraded from boilerplate to a material deal term. It is now worth insisting from the terms stage that the grantor retains approval rights over AI training and output licensing. It’s important to make the conversation explicit.
As for AI companies who don’t seek permission from rightsholders, infringement lawsuits (in the interest of full disclosure, including the one our firm filed against Anthropic on behalf of BMG) are already underway. Copyright infringement is still copyright infringement, with the applicable consequences.
Fair Compensation
In determining royalties for AI training uses, the royalty to the artist or songwriter for AI licensing should be the greater of the applicable royalty under the agreement, or a true 50/50 split. Since publishing deals customarily pay out 75%+, the 50/50 approach is more applicable in recording agreements, especially with the majors, where even a 50/50 net profit split is calculated after a distribution fee to the label. This should apply to all unallocated income (e.g., black-box and breakage) without qualifiers (like an eligibility threshold), to the extent relevant. Also, royalties should flow through without applying to any unrecouped balances.
This is a new medium with untested ground, and it’s only right to let creators in on the upside as an equal partner — and incentivize them to participate.
Relatedly, vocal rights, which are very valuable for the output function of any AI offering, should be entirely reserved to the artist. Whatever rights a label holds to an artist’s existing recordings—even with exclusivity on recording rights in term—are completely different from AI-generated replications of the artist’s voice on new material. Any attempt to wrap in these rights by a label is a severe overreach. The voice of an artist belongs to that artist, period.
Prepare for Contingencies
There has been a lot of discussion about the concept of AI generation living on a given platform within a “walled garden.” AI offerings such as Udio have reportedly taken advanced technological measures to make sure generated content only lives on-platform through various digital rights management methods, while promising that if those fail, leaks will promptly be removed.
Many artist representatives have trouble believing that the beast will be 100% containable. So, we need to consider what happens when we find ourselves in a classic music industry “whack-a-mole” situation, where enforcement is futile. It’s necessary to somewhat accept this possibility to fully commit to AI licensing, so let’s at the very least establish systems to monetize and track everything if Frankenstein’s monster escapes the “walled garden.”
Another consequence of the “walled garden” AI offerings is that without the ability to distribute more widely, these are inherently capped in their broader cultural reach. The next frontier will be a legitimate licensing structure for programs, such as Suno, which explicitly aspire to be used for making music for commercial exploitation.
Transparency
An essential step to adequate compensation—as well as rights enforcement—is an understanding of full transparency between the AI company and rightsholders and between rightsholder companies and their writers and artists. It’s necessary to have systems for precisely tracking and analyzing AI usage at every step of the process.
This will become especially important if Suno and other programs intended to make commercially exploitable music want to achieve legitimacy (Suno has a deal in place with Warner Music Group, but discussions with UMG have stalled). With this type of program, extreme transparency will be needed in every corner and detail of any AI generation.
Whether consumers of AI commercial music deserve transparency is another philosophical debate. Songwriter attribution, at the least, would offer a window into how these programs are used (like any derivative composition, AI generated compositions will need to reflect the original authors’ ownership). But a word of warning to anyone who underestimates their audience: fans can smell inauthenticity and spend their money accordingly.
Looking to the Future
A lot of these concepts are embodied in the principles set forth by the Music Artists Coalition (MAC), an important creator advocacy group, in their press releases and discussions on AI licensing developments. They distill it to three essential concepts: clarity, compensation and consent. MAC and Manatt are leading the charge on attempts to establish a consensus around ethical AI use from an artist and songwriter rights perspective.
On a personal level, as a lifelong music fan, I don’t fear the onslaught of AI to the detriment of music creators. Music is so impactful because it is human — full of pain, joy and love that can only come from the heart. AI can try to replicate that feeling that hits you in the gut when you hear a great song for the first (or hundredth) time, but still falls short. Human inspiration is still needed to push boundaries and come up with new ideas. That is why we need to ensure that creators keep creating, inspiring and fueling music creation. The soul of our industry depends on it.Music Business Worldwide
