Who Pays When AI Steals Your Voice?

What do we think about AI-generated content? No, let's be more specific. How do we feel about artificial intelligence (AI) emulating your favourite artist? Yes, this month, I'm going in on the implications of plug-Ins that mimic the voices, vocal styling, and even their cadences and what that means for intellectual property rights. By now, you've probably heard of the David Guetta track featuring an AI 'Eminem' or the viral song, by a social media account named Ghostwriter, with an AI 'Drake' and AI 'The Weeknd'. But I want to look broader than that. The journey to understanding my feelings about AI emulation has been long, and I'll begin on YouTube.

Spectre does something exciting with the 'But Just My Voice' series on his YouTube page of rap covers. He recreates complete albums using just the human voice with the help of a few collaborators. I can't call it 'beatboxing' since it feels like so much more than that. Undoubtedly, this is an art in the legacy of Doug E Fresh, Rahzel and Beardyman, but this work pushes it further. Rappers' voices, drums, instruments, samples, and sonic effects are reinterpreted by the human voice. The complete tracks are a homage to the artist's body of work. In many ways, it's a cover version. I'm also catching up on ITVX's Starstruck, where singers perform tributes to well-known musicians. It's primetime television. Entertainment by impersonating others has been this way for a while - remember Stars In Their Eyes from the 1990s?

We're entirely comfortable with Ariana Grande or Jay Pharoah doing their imperceptible take on their fellow artists by mimicking them. However, using technology, specifically AI, to emulate musicians' voices is also becoming one of the most controversial conversations in our industry. Jay-Z's longtime sound engineer, Young Guru, recently posted a viral Instagram video. It's a demo track by AllttA (American rapper Mr J Medeiros and French producer 20 Syl) and features 'Jay-Z', where an AI plug-in generated the rapper's voice. Yes, a Jay-Z song without Jay-Z, and it's flawless. Utterly indistinguishable from the real thing. The David Guetta song featured a couple of lines from a fake Eminem. The AI 'Drake' track sounded unfinished to a trained ear - a reasonable demo. But this is class. Lyrically, production, attitude and sound are precisely Jay-Z. Young Guru typed:

"I've been trying to tell everyone that this is where we are now with AI. For some reason, this one got everyone's attention. So what do we do? On one hand, I'm well aware that you can't stop technology. Once the genie is out of the box you can't put him back in. On the other hand, we have to protect the rights of the artist."

While some argue that this technology opens up new creative possibilities for musicians, there are also concerns about the ethics of using AI to replicate the voices and styles of musicians without their consent. For many, it raises serious questions about ownership and the right to control one's likeness and voice. There have already been legal disputes over the use of musicians' voices, alive or dead, in commercials and other forms of media. Those disputes are often settled based on the access to or use of the physical media, e.g. who owns the master tapes or video on which the original voice and the likeness are captured. It's how clearing a sample works. It's more than the creative work itself. It is the right to use the creative work as it's captured on a medium.

As technology advances, so does the ability of AI to mimic human voices, styles, and cadence. Unfortunately, AI has become so expert at what it does that we don't know how it does it. A few years ago, I produced a podcast with technologist Geoff White. Listen to our episode on AI in Health to be amazed by how well AI diagnoses cancers, but the medical experts are baffled about what it's doing. The same kind of technology has led to the development of AI plug-ins that emulate musicians' singing and rapping styles, even after their death. It learns a pattern based on the data (audio, video, photos, numbers, etc.) and reproduces it as it believes it accurately represents what it has been fed based on a new request. While some view this as an exciting technological advancement, others argue that it raises serious ethical and legal concerns.

What if AI could allow us to hear artists we would never hear again? Remember the recent use of green screen techniques to 'resurrect' the late rapper 2Pac in a posthumous performance at the Coachella Valley Music and Arts Festival in 2012? Then, it was a hologram projected onto the stage. Today, AI software would allow 2Pac to 'perform' alongside live musicians and respond in real-time. Imagine seeing Amy Winehouse, Jimi Hendrix, Bob Marley, The Notorious BIG and Aaliyah on stage again? Or should we accept that once an artist has left us, that is it? No more music, no more performances. No more voice.

In a 2017 interview with Rolling Stone, the estate of the late musician Prince stated that they would not allow his music to be used in any form of advertising, stating that "no permission was granted to the producers of the Halftime Show to use Prince's name, likeness, or music." This raises the question of whether or not musicians or their estates have the right to equally control the use of their voices and likeness in AI plug-ins.

On the other hand, optimistic creators using AI technology argue that the advantage it gives can be used to extend the legacy of deceased musicians, allowing their music to continue to be enjoyed by future generations. I'm focusing on musicians no longer with us because that is the clear argument for AI. If there was no other way to get that voice into a studio. I rest my...

Hang on. There is one more perspective. An intriguing one that suggests that AI plug-ins can also be used as a tool for aspiring musicians to learn from the styles and techniques alongside their favourite artists - like the bunny in a greyhound race or a virtual sparring partner. Let's call it the 'Virtual Mentoring' defence. Using an existing artist as your creative training wheels, or if songwriters used AI emulators to 'sing' a guide vocal, then the creators are not commercially exploiting someone else. It's more manageable for lawyers.

There are already legal precedents for protecting the commercial use of our likeness or the media it is captured on. AI is no different. Digital Media lawyer Cliff Fluet who has been advising on Creative AI for the last decade, believes “we’re now understanding what’s really valuable about artists and music – in the past, it was all about sound recordings … now the value is in the song, the artists’ name and likeness and now their voice ..?these new areas of IP need to be understood and protected before they can be licensed.” If it sounds like Jay-Z and you try to profit from it without Jay-Z, the lawyers will come for you. Unless it's covered by the conditions of fair dealing (fair use in the US).?

It's all subjective, however. "The government is not smart enough to regulate AI" - a line that struck me as I listened to composer benbrick’s Future Shock audio essay about Artificial Intelligence. It reminds me that the ethics around AI isn't so much about whether AI is good or bad. It's not always clear cut, black or white. There is something else that we expect from our artists - authenticity. A sense of integrity we associate with what we know about them, what they stand for and what we think they believe. It's the same with Deep Fakes when other people's faces are mapped onto existing photos or videos. We know it's unlikely to see a video of Dame Judi Dench, for example, doing something inappropriate with her clothes off (I'll leave it there) if we saw a video purporting to be so. We immediately get that it's a spoof or some creative experiment, but perceptively not the real Judi Dench.

Context is everything. Faking voices have the same issue. An AI Justin Bieber could make a song in support of child grooming gangs. It is extreme but possible. In 2020, at the height of the Stormzy and Wiley beef, we created this parody video of what a collaboration could look like. It took 46 hours to put together and only got as far as our 823 social media followers. The AI Drake song 'Heart on My Sleeve' was streamed 15 million times on TikTok in the same amount of time. What compounds the conversation around AI is that today everything moves quickly. What we believe we're seeing or hearing can be around the world and reasonably monetised before it can be authenticated.

To follow benbrick's lead - it will make sense when you listen to Future Shock - I'll predict how this could play out. The lesson learned from Napster is that the industry eventually joins them when they can't beat them. In the meantime, just like with the industry that grew around sampling culture, the clever lawyers and the rights owners will find a way to come after the creators using AI, either by pursuing claims when a track sounds like their client (e.g. Robin Thicke vs Marvin Gaye's estate) or?claims that the generated voice is generated on material that they own, e.g. acapella, lyrics, etc. It's easier to track. The challenge for creators is that as it's all data, there's a 'paper' trail of the file names you use.

So what could an official AI emulator look like? It's a simple idea. Imagine a world where your favourite singer or rapper's voices are licensed like production sample packs, like Splice. For instance, you can buy a Drake AI emulator plug-in, but if you want to use it on a commercial track, you must deal with Sony/ATV Music, which owns the publishing.

There could be the Death Row vocal pack which could feature Dr Dre, Snoop Dogg and 2Pac, as well as Warren G, D.O.C. and MC Hammer. How about the Bad Boy vocal pack with Notorious BIG and Ma$e, which will also give you access to an AI emulator based on the original recordings of G Dep, The Lox, Shine, Black Rob, Loon and Craig Mack?

There could be two packs for Michael Jackson's AI voice - Motown recordings for the early years and a CBS one from 'Off The Wall' recordings onward. How about the Spice Girls voice pack, which would include individual modes for the solo artists or a group mode emulating their harmonies? Producers could buy these AI plug-ins outright or subscribe to a usage plan, like for Getty archives and Shutterstock.com.

Creators could use the AI voices for mashups, remixes or even original tracks. However, if they want to release a track commercially, there would be a licence fee to negotiate and points on publishing. There would be limitations on additional earnings from the AI-voiced songs because the producer couldn't promote or perform the track (where big money in music is made). Please don't shoot me. It could be a re-issue of recordings in a hyper-digital form. Instead of getting alternative or extended mixes with instrumentals and acapella, you now get the building blocks for an entirely new creation.

Whether or not using AI to mimic musicians' voices and styles is ethical remains to be answered. Being possible and being ethical are not the same argument. Some argue that this technology crosses a line when using an actual individual's likeness and voice. While using AI to mimic musicians' voices, styles, and cadence is an exciting technological advancement, it raises serious concerns because you are taking away their ability to earn money from the image and talent they have developed, maintained and built a fanbase around. A completely original, computer-generated voice is less problematic as long as it is not stealing from or biting another artist, as the old Hip Hop crews say.

To quote the West Country poets, Goldie Lookin' Chain, "Guns don't kill people, rappers do?". AI technology is ultimately a tool. A power one but simply a tool. Like the gun. The intention of the person who presses the button defines the use, legality and morality of what comes out of your headphones. After all, when the lawyers come knocking, they are coming for the content creator and not the plug-in. As long as we arm ourselves with the rules around copyright, intellectual property and protected assets, we'll be fine - until the robots take over.

Want to talk more about AI and IP? Don’t hesitate to drop me a line on LinkedIn.

Want more stories like this??

Subscribe to our Newsletter.

###