The combination of quality, access and user appetite for AI-generated music has forced major labels to scramble to form a strategy to protect their investments. This has taken several forms, from threatening the developers of AI systems against using their recordings to train algorithms to exerting pressure on streaming services and lobbying lawmakers. Reminiscent of the early days of Napster and the proliferation of peer-to-peer music-sharing services, with the AI genie out of the bottle, will it be too little, too late?

There has been a keen focus by music rightsholders on both the so-called “input“ and “output“ stages of AI systems. On the input side, major labels have been unequivocal in warning developers of AI systems that they intend to take action against anyone who trains algorithms using their recordings. Furthermore, in March 2023, more than 40 organizations from across the music, sports, film, journalism, and other industries, including the Recording Industry of America, SAG-AFTRA, SESAC, and SoundExchange, launched the Human Artistry Campaign (HAC). This campaign aims to ensure that the use of copyrighted works by AI developers and users requires authorization and licensing and that governments do not allow for copyright exemptions that would permit the exploitation of copyrighted works for AI training without permission or compensation. Significantly, HAC’s membership base increased by 50% in just one month from its launch, demonstrating the level of concern across creative industries surrounding AI models. In the section, Why the AI sector has a training problem, we explore in further detail the copyright implications of training and territorial differences in approaches as governments race to attract AI talent and investment.

The output side is no less complex. We explore further in the IP: Copyright section if works solely produced by AI (as opposed to AI assisting human creators) will receive copyright protection. Further, the emergence of AI sound-alikes, where tools are used to synthesize voices to imitate famous artists, has been increasingly visible on streaming services and social media. It has triggered a spirited debate involving artists, industry executives, fans, and lawyers. AI-generated sound-alikes are typically not created using samples or snippets of existing recordings. Rather, these sound-alike recordings are generated independently using AI tools trained to recreate a specific voice. This reality renders ownership of the original sound recording’s copyright somewhat powerless, as the newly recorded material is not a direct copy of an existing work. In response to this gap in copyright law, consideration must be given to the rights of publicity, which can safeguard a celebrity’s name, image, likeness, voice, and other unique personal attributes from unauthorized commercial use. However, the legal landscape is diverse and often complicated, as we explore further in the Geopolitics of AI section.

The surge of AI-generated sound-alikes presents an array of legal, ethical, and practical challenges. This situation echoes the issues surrounding sync rights and short-form user-generated content platforms. With this new wave of technology, will the music industry persist in an endless cycle of takedown requests, apply pressure on digital service providers with new restrictions in licensing agreements, and push for legal reform, or will it adapt to and find ways to profit from this new wave of creative output?

One approach that some have suggested is that artists could actively engage with this technology themselves, creating their own AI voice models for use in various commercial contexts. This could open up new opportunities for monetization, allowing artists (or even investors in the space) to license their voices for use in anything from virtual concerts to AI-created fan experiences. Indeed, some artists and estates have already begun exploring these avenues, with the holographic performances of deceased artists like Roy Orbison and Tupac Shakur as well as live artists, like ABBA, offering glimpses of what could be possible.

However, such a model would likely require the placement of significant legal protections to ensure artists retain control over their voice and can prevent its misuse. It would also raise additional questions about how revenue would be distributed. Would revenue from AI-generated performances be treated like royalties from cover versions, or more like revenue from merchandise or brand endorsements?

As AI continues to evolve, it will undoubtedly transform the music industry in both positive and challenging ways. The key will be finding a balance that allows for innovation while respecting the rights and contributions of human artists. The future may hold a mix of AI-generated and human-crafted music, with collaborations between artists and algorithms becoming the norm. Ultimately, the human element of creativity and expression will always be at the heart of music, and AI will serve as a powerful tool to amplify and expand artistic possibilities.