MBW Views is a series of exclusive op/eds from eminent music industry people… with something to say. The following op/ed comes from Ran Geffen Levy, Founder of OG.studio, which provides insights to music tech start-ups, companies and VCs. He is also the CEO of Amusica Song Management in Israel.
A quiet land grab is happening in the music business, and if publishers and songwriters don’t act fast, they’ll miss the moment entirely.
This isn’t just a theory. This week, a judge in California dismissed Universal et al’s motion to block Anthropic from using copyrighted song lyrics to train its AI, saying the publishers hadn’t shown “irreparable harm”. While the judge recently denied a preliminary injunction on narrow procedural grounds, she explicitly stated that fair use remains an unsettled question and acknowledged the emergence of a licensing market for AI training. The publishers remain confident and are vigorously pursuing monetary damages.
This ruling wasn’t just a legal setback; it was a warning. If songwriters and publishers don’t define their role in this economy, the courts won’t do it for them. This vigor should be directed also to other stake holders in the music industry. Songwriters, composers and music publishing communities must demand their fair share.
AI companies are moving quickly, building tools that generate songs, clone voices, and mimic styles. Master owners are already in the room. They’re licensing recordings for training and, in many cases, walking away with something far more valuable than a training fee or output royalty: They’re gaining equity in the AI start-ups themselves.
So, here’s the question no one’s asking out loud: Where are the songwriters in these deals?
And here’s another: What about the billions of dollars already invested into music copyrights by major financial players, many of whom control both the publisher’s and the writer’s share? Over the past few months, I’ve spoken with some of them: leaders who’ve acquired catalogs [at] scale and are closely involved in the work of collective rights organizations. Their insight and leadership could be essential in shaping a new framework, one that not only preserves the value of their investments but helps define the next era of rights and revenue in music.
If AI is the next economic layer of music, and it’s being built on massive training datasets, then why are the owners of the compositions, the melody, the lyrics, the emotion, being left out of the foundational infrastructure?
This is the new value gap. It’s not about big tech building value from recordings and compositions anymore. It’s about who in the music industry owns the pipes in the AI age, and who’s treated as nothing more than source material.
And if publishers and writers don’t move fast, they won’t just be excluded from the profits; they’ll be excluded from the decisions that shape what music becomes.
Here’s a clear plan of action that will enable publishers and songwriters to move from the sidelines to the center and to future-proof their role in the AI-driven music economy.
1. Demand Transparency: No More Backdoor Deals
Publishers must start by asking master owners the hard questions: What deals have been made? Who are the partners? And most importantly, under what authority? Master owners cannot cut AI training without a song. Therefore, they should not be able to cut those deals without the written approval of the songwriters and their publishers. That needs to be said plainly and enforced collectively.
Yes, the industry is demanding broader training disclosures from tech companies. But on this front, we must walk side by side with the master owners, not behind them. The core issue is this: every time a recording is used in training, and that recording includes a copyrighted composition, the songwriter and publisher must have a seat at the table, invited or not.
2. The New Rate Plan: Introduce a fair royalty model based on the type of data used
Not all AI training is the same. We must distinguish between models trained on compositions, instrumental recordings, and vocal performances. And compensation must reflect those differences.
Here’s a proposed framework:
- 100% to publishers when AI is trained on synthetic recordings derived purely from their compositions.
- 50/50 between publishers and master owners for training on instrumental recordings.
- Equal three-way split (33/33/33) between publishers, master owners, and performers when vocal recordings are used when the voice is not tied to a specific, named artist.
In what’s become known as the Grimes model, where a creator uses a specific artist’s voice to generate new music, the artist receives 50% of the master royalties, while the new composition, if original, is fully owned by the writer. That sets a precedent: if a voice powers the output, it gets paid.
“AI doesn’t draw inspiration, it draws data.”
And that logic extends further. If a human creator uses a dataset trained on a specific writer or producer’s style, that contributor should be recognised and compensated as well. Style has value. Influence has value. And AI doesn’t draw inspiration, it draws data.
Separately, when human creators use AI as co-writers or tools, legally retaining 100% of the rights under current US Copyright Office guidance, a data usage fee must still be paid to the rights holders whose works trained the models. Whether it’s compositions, recordings, or voices, no one’s contributions should be used for free.
Sometimes these models might seem complicated or impractical for many reasons I won’t address here, but I can assure you that technological solutions already exist to enforce all of them. The real challenge now is who will step up first and build a comprehensive dataset fully owned by creators and publishers—no masters needed.
What’s often overlooked is just how much leverage publishers already have, if they choose to use it. In recent years, billions of dollars have been invested in music copyrights, particularly in publishing, by major financial institutions, pension funds, and private equity. These are not passive investors. Some of them also hold stakes in collective rights organizations and performance rights societies. They are shaping the infrastructure of rights, not just funding the catalogs. This isn’t just about the music industry anymore. It’s about a wider financial ecosystem that has the resources, scale, and strategic vision to influence how rights will be valued and governed in the AI era. If they choose to act, they won’t be asking for a seat at the table. They’ll be shaping the table itself.
One final warning: The industry must resist blanket license solutions. While efficient on the surface, they’d reinforce existing inequalities, favor dominant catalogs, and marginalize independent writers and smaller funds holding partial stakes. The AI era is our chance to fix systemic imbalances, not amplify them. Fair attribution must be a core principle from the start, not an afterthought.
3. Strategic Clusters: From Passive Licensing to Active Participation
The current music rights system was built around sync, mechanical, and performance rights. But AI changes everything, and we need to change how we organize and monetize rights.
Today’s songwriters often work in groups. That’s not new. What’s new is the opportunity to formalize those collaborations into strategic clusters: rights bundles that include compositions, voice models, stylistic data, synthetic performances, and even visual or narrative personas.
These clusters can be licensed, tracked, monetized, and even leveraged for equity, just like master owners are doing today. While AI poses real risks to traditional songwriting models, it also opens the door for songwriters to become more autonomous. They are no longer entirely dependent on external performers to bring their work to life. By building AI-ready catalogs, songwriters can shape how music is made, sold, and experienced in the future, with greater creative freedom and control.
But that’s only part of the equation. These clusters will also require a new rights infrastructure.
4. Collective Rights Society 2.0: Building Tomorrow’s Framework
The traditional collective management systems were not built to track synthetic recordings, AI collaborations, or digital personas. That’s why it’s time to begin designing what comes next: a forward-facing framework that reflects the reality of AI-era creativity.
This new model won’t just include publishers and songwriters. It must bring in model creators, data contributors, voice owners, and new kinds of participants we’re only beginning to imagine. The players are changing. The rules must change with them.
It’s also important to recognize that rights management and regulation will evolve differently across territories. What applies in the United States today may look entirely different in Europe, Asia, or Africa tomorrow. Some regions will move faster than others in defining what constitutes fair use, consent, or compensation for training data.
Traditional collective societies might not adapt voluntarily to AI-era needs. But several are now owned by investors with strong incentives to future-proof royalty collection. The technology needed for this already exists. The real question is whether the industry will proactively implement these solutions or wait until external pressures force less favorable terms.
The Billon Dollar Question
As of now, nothing has been universally decided. That uncertainty is precisely why publishers, rights holders, and emerging digital actors must collaborate, not wait, to shape the next international framework together.
This is a rare moment to build something better. The industry doesn’t just need protection. It needs vision. It’s time for the players who care about music’s future to step forward and lead. This is no longer just about protecting the past. It’s about building the future. The framework is here: transparency, fair royalty models, strategic clusters, and a new vision for rights management.
Here is the billion-dollar question: every income stream described here is still theoretical. Billions have been invested in music rights based on assumptions tied to predictable copyright income in a pre-AI world. Right now, there’s no legal guarantee for revenue from AI training, placing these investments at risk. Publishers and songwriters must urgently prioritize addressing the new value gap. Without immediate action to bridge theoretical value and legally protected revenue streams, a significant correction in copyright valuations is inevitable, putting billions in capital at stake.
What happens next will depend on who’s willing to act, and who’s willing to lead. The time for songwriters and publishers to claim their rightful place in the AI music economy is now, before the door closes for good.
Disclaimer
I’m a big fan of AI and I regularly use it, including to fine-tune and articulate my ideas for this op-ed. I believe in giving credit: ChatGPT helped me sharpen my thoughts, while Claude acted as the devil’s advocate, identifying weak points and tying loose ends. I love AI and genuinely believe it empowers humans to create in ways we’ve never experienced before.
Music Business Worldwide