AI and Likeness: Let’s Ask Michael Nash

As EVP/chief digital officer at UMG, Michael Nash is as qualified as anyone to comment on all aspects of AI and music—its perils, its promise and its peculiarities. As such, he is integral to label group boss Sir Lucian Grainge’s aggressive, wide-ranging campaign to protect artists and copyrights online. Why he decided to waste his valuable time answering our fool questions is more difficult to explain.

What are some of the forms unauthorized AI use of an artist’s likeness can take?

It’s important to start with the basic fact that the vast majority of uses are totally unauthorized and created with unethical, unlicensed AI. We see cloned voices used for impersonated tracks and covers, fake leaks passed off as unreleased music and AI tracks on real or lookalike artist profiles, typically as vehicles for streaming fraud.

We also see fake ads, political endorsements and deeply offensive content, including sexualized imagery and fabricated scenes of violence. Some of it attacks an artist’s identity directly.

We’re committed to protecting the artist, even when a deepfake doesn’t involve the recording or composition rights we directly represent. An artist’s identity is personal. It’s also their livelihood.

How do you address it? Do you feel it’s a sort of whack-a-mole?

I wouldn’t say “whack-a-mole,” but the scale of the challenge is certainly staggering. In the last few weeks, we sent around 75,000 takedown notices for deepfakes, which had nearly 500 million plays. We work with the most popular and recognizable artists in music, and they can therefore be among the biggest targets. The integrity of their brand is important to the entire ecosystem: Real career artists are what bring fans to these platforms.

Our response requires automated detection, advanced tools and platform action. We combine our own monitoring with specialized technology and specialist expertise to identify different kinds of abuse. We track how deepfakes are made and spread, but the tactics of bad actors can change as fast as the technology. We work with platforms to take them down, using each case to sharpen and evolve our response.

We work with the most popular and recognizable artists in music, and they can therefore be among the biggest targets. The integrity of their brand is important to the entire ecosystem: real career artists are what bring fans to these platforms.

Takedowns are an important safety net, but they’re not enough. As Lucian said on our latest earnings call, the flood of AI slop, fraud and abusive content is accelerating. Technical intervention and structural guardrails have to keep up, while business policies and content-management practices need to adapt. Artists and their music create enormous value for streaming platforms. Protecting artists is in a platform’s own interest. Partners need to work with us on solutions that move faster and work at scale.

Are your artists coming to you about unauthorized AI uses of their likeness?

Yes, increasingly. Our monitoring catches cases, and artists and their teams bring us what they find, too. A deepfake can reach millions of streams before the artist even knows it exists. Fans may well be fooled, and the artist is left dealing with the fallout. That’s why clear labeling and fast action matter.

History shows us that music is often at the forefront, encountering and addressing the impact of new technology. With AI, recording artists may be among the first targets, but nobody is immune. Imagine a deepfake of you going viral, putting words in your mouth that violate your values. That’s already a reality for artists. We need guardrails and legislation to protect everyone. That includes a federal right of publicity in the U.S. as a top priority.

Fans want to interact with artists by creating something new from elements that may fall under the rubric of likeness—image, sound, style. How do you navigate this?

It’s helpful to know we’ve been here before: We found ways to license music and related content in user-generated videos, implement effective content management and compensate artists. We can achieve the same holistic balance with AI.

A deepfake can reach millions of streams before the artist even knows it exists. Fans may well be fooled, and the artist is left dealing with the fallout. That’s why clear labeling and fast action matter.

UMG research across 13 representative global markets found real consumer interest in using AI to remix, reinterpret or personalize songs. But interest isn’t permission. The opportunity is to turn that demand into artist-approved creativity.

This is what distinguishes our strategy—we’re embracing artist-centric AI innovation, not just enabling a free-for-all that disestablishes artists’ rights and undermines their interests. Our agreements with Hook, Klay, Udio, Spotify and ElevenLabs are designed to give fans authorized places to personalize, customize and interact deeply with officially opted-in artists’ work. There are guardrails that protect artists and a licensing model that ensures appropriate compensation for the creative community. The test is simple: It has to work for artists and benefit the wider ecosystem, too.

What kind of uses are artists OK with, and what uses might they be prepared to allow if they were monetizable on an automated basis?

For an artist’s name, image, voice and likeness, the baseline is consent, compensation and control. Artists are more likely to welcome uses that give them agency and deepen their connection with fans.

The possibilities are broad, and the best uses often emerge through dialogue between artist and fan, a process of co-creation. With an artist’s approval, fans might translate a song into another language to better understand its meaning, reimagine a music video or make an acoustic version for a different setting. This superfan engagement can even spark ideas for future artist projects. Licensing can help make these uses work at scale, but the economics must reflect the value the artist brings.