Artificial Intelligence is the protagonist of numerous news stories on a daily basis: hailed as a hero in the healthcare industry, AI is highly criticized and considered a dark evil power and a villain in the creative arts. We recently analysed the exploits of Artificial Intelligence in the realm of fashion, but there is another area where it has generated significant controversy - the world of music.
There are generators that quickly allow you to create music, such as AIVA (used to generate the tracks featured in this post) that allows you to compose a variety of tracks in a wide range of genres and moods, from Cinematic Orchestra to Cyberpunk, Drum'n'Bass, Trip Hop, Lo-fi and so on. The length depends from the subscription and only the Pro Monthly package ($49 monthly) allows you to own the complete copyright of what you created). At the moment, the sounds produced are not extremely original and quite often you don't get what you wanted, but you can spend some time playing on it and come up with some experimental tracks for fun.
Things are moving fast, though in AI-generated music: a few days ago, Grimes announced that she welcomed musicians who wanted to create new tracks with Artificial Intelligence using her voice. On Twitter the Canadian singer, whose real name is Claire Boucher, stated: "I think it's cool to be fused w [sic] a machine and I like the idea of open sourcing all art and killing copyright."
In another tweet Grimes announced, "Ill split 50% royalties on any successful AI generated song that uses my voice. Same deal as I would with any artist i collab [sic] with. Feel free to use my voice without penalty. I have no label and no legal bindings."
The singer went on to highlight that she will not accept songs with violent or offensive contents, "may do copyright takedowns ONLY for rly rly [sic] toxic lyrics w grimes voice (…) Rly don't like to do a rule but don't wanna be responsible for a Nazi anthem unless it's somehow in jest a la producers [sic] I guess."
Grimes is not new to using AI: in 2020, she collaborated with music company Endel to create an AI-generated lullaby for her son X Æ A-12 (his father is SpaceX and Tesla founder Elon Musk).
Compared to other artists who have expressed their displeasure with the use of Artificial Intelligence in generating music, Grimes appears to have a more nonchalant attitude towards it. However, she may be an exception: we have seen in previous posts how AI is being trained by users to create credible fashion collections, but it is also being utilized to produce artificial voices that perfectly reproduce famous artists, something that has stirred up indignation among many musicians, including Drake. Most artists seem gravely concerned about this issue and consider AI as a threat to the industry.
Recently, Universal Music Group (UMG) asked TikTok, YouTube, and Spotify to take down a song titled "Heart On My Sleeve," by anonymous creator ghostwriter. The song - that went viral before being taken down from platforms including Apple, Soundcloud, and Bandcamp - contained vocals produced by AI that imitated the voices of their musicians Drake and the Weeknd. UMG compared AI-generated music to deep fakes and fraudulent practices and stated this amounted to a violation of copyrights and deprived artists of their rightful compensation.
Drake also didn't seem very flattered when another artificial version of his voice was used in a cover of Ice Spice's song "Munch", released with a fake verse by him. In an Instagram story Drake wrote: "This is the final straw AI". But there have been more AI generated tracks recently, featuring the voices of Rihanna, Kanye West and Travis Scott, just to mention a few ones.
In April, in an effort to protect their copyrighted songs from being used by Artificial Intelligence platforms, UMG sent a request via email to streaming giants like Spotify and Apple Music to block the access of AI systems to the melodies and lyrics of their songs. Yet already a few months ago, in October 2022, the Recording Industry Association of America (RIAA) had already issued a warning about numerous AI enterprises were violating copyright laws by using copyrighted music to train their AI algorithms.
However, there are several legal issues to consider, first of all, technically, a voice cannot be copyrighted but famous individuals are safeguarded by their right of publicity, protecting them from privacy invasion and appropriation. There have indeed been cases in which artists proved they have a right to their voices: in a 1998 Ford commercial, a Bette Midler's backup singer was hired to imitate her voice when the actress declined the job. Midler sued for copyright infringement and won (Midler v. Ford Motor Co.).
In much the same way, in 1988 Frito-Lay used an impersonator of Tom Waits in a Doritos commercial, the singer sued for voice misappropriation under California law and false endorsement under the Lanham Act and the jury found in Waits' favor, awarding him compensatory damages and punitive damages for voice misappropriation (Waits v. Frito-Lay, Inc.).
Yet in 2004, in the case of Butler v. Target Corp. the final verdict was different. Ishmael Butler and Maryann Vieira penned the jazz/hip-hop music composition "Rebirth of Slick (Cool like Dat)" recorded by Butler, Vieira and Craig Irving under the name Digable Planets. The song was used in a commercial multimedia campaign for Target Stores that employed altered versions of the song with slogans such as "Jeans Like That," "Denim Like That," "Shoes Like That" and the like. The court determined that while the lyrics of a song are subject to copyright protection, the unique voice used to perform them wasn't. This is because spoken words are not considered fixed in the same way as recorded music, and therefore cannot be protected under copyright law.
That said, is an AI-generated rendition of an artist's voice to be considered a violation of copyright law or an imitation of that voice. After all, would you be able to sue a performer who perfectly imitates Drake? No, you can potentially sue an impersonator for violating the celebrity's "right of publicity", but not for mimicking a celebrity's voice without their consent.
Besides, the legal standing on whether a piece of art created by a human that incorporates AI components can be protected by copyright remains ambiguous. In ghostwriter's case we don't know which parts of the track they created using AI.
We do know, though, that, for what regards AI-generated materials, in February the U.S. Copyright Office (USCO) stated that, to be registered, an artwork must be the product of human authorship.
At the same time, in new guidelines released on 16th March 2023, the US Copyright Office highlighted that there may be AI-generated material that may contain sufficient human authorship to support a copyright claim. For example, a human may select or arrange AI-generated material in a sufficiently creative way that "the resulting work as a whole constitutes an original work of authorship." Or an artist may modify material originally generated by AI technology to such a degree that the modifications meet the standard for copyright protection. In these cases, copyright will only protect the human-authored aspects of the work, which are "independent of" and do "not affect" the copyright status of the AI-generated material itself.
Therefore, copyright determinations will need to be made on a case-by-case basis, which may mean that ghostwriter may still have a right to claim the copyright of their track if it contained "sufficient human authorship".
Another issue arises when it comes to stopping the use of artists' music to train AI applications. Similar to AI text-to-image applications that learn from vast amounts of data, AI music production relies on being trained with music and audio tracks, sometimes even entire libraries of music.
UMG claim that the use of their artists' music for training generative AI was a clear infringement of copyright law. Nevertheless, there is no proper law about this, so Universal's position remains untested in a court of law. At the same time, UMG has admitted to use AI in their music production (we do not know in which ways - do they have their own system and do they feed it with hours of music recorded by anonymous artists? Do they feed the AI music from their own libraries?), yet they appear to be opposed to individuals using AI to recreate their own artist's music (obviously for money reasons...).
One of the challenges with the use of AI in creating art is that there are currently no regulations in place to restrict what AI can and cannot be trained on. UMG is free to send cease-and-desist letters, but it is unlikely that AI services will stop accessing copyrighted materials unless legal boundaries are established.
There are a few ways that music sites can protect their artists from being copied by Artificial Intelligence. One method may be using a form of digital watermarking technology to embed unique identifiers into the music file or apply some kind of limitation just like a robots.txt file that controls which URLs search engine crawlers can access. Yet there is no guarantee that these methods would be effective at all.
This rapid production of music and voice through Artificial Intelligence and the ever-evolving technology behind AI present a challenge to major record labels.
There are several issues, questions, doubts and dilemmas regarding AI: will it damage artists and how can certain industries and talents survive? Such questions may go unanswered for if Artificial Intelligence doesn't get regulated in some way: the main problem remains the fact that the technology is already available to be used in the mainstream and people are employing it, but not rules exist.
While the possibilities of AI remain fascinating, the technology needs to be regulated in some ways. One solution in these cases would be banning AI companies using copyrighted works to train their models without permission. But there are other ways to regulate it and they may come from artists: Grimes' offer isn't the best, actually, for one main reason, she isn't licensing her voice to users to make a track, but she's waiting for users to make the music, to then claim half the royalties, which sounds a bit bizarre for somebody claiming she wants to "kill copyright".
Other artists have suggested better options including an alternative AI identity: Holly Herndon has been exploring the possibilities of AI in music for several years with her collaborator Mathew Dryhurst using technologies such as AI, Web 3.0, and blockchain.
Together they created the virtual persona Holly+ using deepfake technology. People can experiment with Holly+ for free, but Herndon also offers commercial licenses while ensuring she receives a share of the profits through a decentralized protocol. Last November Holly+ covered Dolly Parton's "Jolene," using a modified score of the original song composed of new tonal harmonies that was fed into Holly+ and then generated in Herndon's voice.
In 2022, Dryhurst and Herndon together with Jordan Meyer and Patrick Hoepner founded Spawning, a project that enables artists to have control over the use of their work in machine learning-generated media. With Spawning, artists can choose to opt in or out of the datasets used by AI art generators to create their compositions (the team also created haveibeentrained.com, a website where artists can search the LAION-5b dataset of nearly 5.8 billion images used to train AI models Stable Diffusion and Midjourney - if an artist discovers their work in the dataset, they can choose whether to continue or terminate the use of their imagery for AI training through Spawning). Spawning's approach has been praised for its potential to inform the European Union's text and data mining policy.
As with any transformative technological advancement, the music industry is currently in a state of flux, grappling with the necessary adjustments and growing pains that come with the integration of AI technology and you can bet that regulating AI in music will be an even more complex and time-consuming task than regulating sampling.
Legally-wise there will indeed be further issues to consider, from artists who will have to start checking also the contents of their contracts in case they include AI clauses and may end up signing away the rights to their virtual selves, to regulating those AI-music generators that allow you to create tracks by uploading a reference file. This practice may indeed be considered as a contributory infringement, the same accusation faced by file-sharing service Napster in the early 2000s.
A claim of contributory and direct infringement against companies stocking up copyrighted materials may actually be the solution for artists and companies who want to take their revenge: Napster was sued by the entire music industry, and eventually collapsed in 2002; LAION functions in a different way for images, but may find itself in the same position (and a few ones have already sued them...).
So, navigating the legal minefield of AI-generated contents and of music in particular is far from straightforward, or, to put it more simply, as ghostwriter announced commenting on YouTube their Drake video, "This is just the beginning." It truly is.
This post was written in collaboration with Nicola Battista, from Italian label Kutmusic, one of the independent artists who sued Napster in the year 2000.
Comments