From streams to prompts: the next royalty question for the music industry

For much of the debate around generative AI and music, the central question has been relatively straightforward: can copyright protected music lawfully be used to train an AI model without permission?

But the market may already be moving on to the next question.

As AI companies begin entering into licensing arrangements with major music rightsholders, the conversation is shifting from whether music should be licensed to what those licences should actually look like and, crucially, how artists and songwriters participate in the value being created.

These questions could prove to be some of the most significant commercial debates facing the music industry over the next few years.

From litigation to licensing

The tension between AI developers and the music industry has been well documented. Rightsholder have argued that copyright works should not simply become free training material for commercial AI products, while AI developers have sought certainty over their ability to access the enormous datasets required to develop their models.

Increasingly, however, we are seeing signs of a licensing market emerging.

Most recently, Suno launched new AI music models following licensing arrangements with Warner Music Group and BMG. Suno has indicated that future products could allow participating artists to opt in to experiences that enable users to create music inspired by them, with artists sharing in the resulting revenue.

What are artists licensing — and how do they share in the value created?

One AI licence may not cover everything.

An artist might agree to let recordings be used for training but refuse permission for AI-generated songs to imitate their voice. Another might approve a digital replica for a specific product but object to their name or likeness being used in marketing. Consent to training, generation and commercial exploitation should not automatically be treated as the same thing.

Among other things, agreements may need to address:

• precisely which recordings, compositions, performances or other materials can be used
• whether an artist’s voice, likeness, name or other identifiable characteristics can be replicated
• the purposes for which material can be used and which AI models can use this
• whether generated outputs can be commercially exploited
• whether consent can be withdrawn and what happens to a model that has already been trained
• restrictions around sensitive, defamatory or reputationally damaging outputs
• transparency and reporting obligations; and
• how the resulting revenue is calculated and shared.

These are not simple copyright questions. They impact contractual control, performers’ rights, data, reputation and ultimately, bargaining power.

Where does the money go?

Then there is the royalty question. The traditional music industry has established structures in place for allocating revenues from recordings, compositions, performances and exploitation.

Generative AI does not necessarily fit neatly within these structures. There may be no conventional “stream” of that artist’s recording at all. So, what is the relevant economic event?

Is value being generated by the underlying recordings used to train the model? By the composition catalogue? By the artist’s performance characteristics? By their voice? Or simply by the commercial value associated with their identity? Potentially it’s a combination of them all, which makes contractual characterisation of AI revenue particularly important.

The UK position

The UK Government recognises that the AI licensing market is still developing, and is considering how to improve transparency and control for rightsholders.

It is also exploring whether additional protection may be needed for digital replicas of a person’s voice or likeness.

This could become increasingly important as licensed AI artist experiences become more common.

The next generation of music deals

The industry has adapted before: from physical sales to downloads, streaming and social platforms.

Generative AI may require another contractual reset. The key question for artists and songwriters may soon be:

“What have I agreed AI can do with my work and how do I share in the value it creates?”

How Hamlins can help

Our Music team deals with a wide range of contentious and non-contentious music matters including recording, publishing and other contracts, intellectual property and copyright issues, distribution, clearance and merchandising agreements and sales of companies and catalogues. Get in touch to find out more.