When AI learns from music, who gets paid?

The music industry has spent the last few years arguing that AI companies should not be allowed to train their models on copyrighted music without permission. Now a new dispute asks a different question: what happens after permission is granted?

Some background

The American Federation of Musicians (“AFM”), one of the world's largest musicians' unions, has sued Universal Music Group and Warner Music Group. The union alleges that the record labels licensed recordings containing performances by its members to AI music companies such as Suno and Udio, but failed to compensate or credit the musicians involved.

Why does this matter?

When we think about music rights, we often focus on the singer, songwriter or record label. However, many recordings also include session musicians whose talent and performances contribute to the final product. The AFM argues that when these recordings are used to train AI systems, this constitutes a "new use" of those performances. If that is correct, the musicians may be entitled to compensation under existing industry agreements.

The interesting IP lesson

This dispute is not really about whether AI companies need licences. The major labels themselves previously argued that licences were required and subsequently entered into licensing arrangements.

Instead, the dispute highlights a question that businesses across many industries will face:

Who should share in the value created when intellectual property is licensed for AI training?

A company may own or control intellectual property rights, but there are often multiple contributors behind the creation of that IP. As AI licensing becomes more common, questions about ownership, consent, attribution and compensation are likely to become increasingly important.

Why should businesses care?

Many companies are rushing to sign AI-related agreements involving data, content, software, images, music and other intellectual property assets.

Before doing so, it is worth asking:

Who actually owns the rights being licensed?
Are there contractual obligations to employees, creators or contributors?
Will revenue generated from AI licences need to be shared?
Have all necessary permissions been obtained?

Failing to answer these questions upfront can create significant legal and commercial risk later.

As AI continues to reshape creative industries, disputes may shift from "Can AI use this?" to "Who should benefit when AI uses this?"

And that may prove to be the more difficult question to answer.

Dr Bernard Dippenaar

Executive | IP & Innovation

bdippenaar@ENSafrica.com