[Guest post] Bad Bunny shoo shooes anyone liking AI-generated song replicating his style and voice – is he right?

The issues with AI-generated content and AI clones are certainly not going away any time soon. The latest concerns Porto Rican sensation Bad Bunny, as Katfriend Despoina Dimitrakopoulou (Plug and Play Tech Center) explains. Over to Despoina:

Bad Bunny shoo shooes anyone liking AI-generated song replicating his style and voice – is he right?

by Despoina Dimitrakopoulou

Recently, the news of reggaeton mega-star Bad Bunny's eloquently put disappointment spread on social media, bringing up interesting questions concerning music creation using AI.

This song called "NostalgIA" (IA standing for “Inteligencia Artificial” in Spanish) was created by account user “FlowGpt”, using the voices of Bad Bunny, Justin Bieber, and Daddy Yankee. The song, and specifically the part of Bad Bunny´s voice and style of lyrics, was reported by his fans to be so much like Bad Bunny that had them fooled. Even singer Bad Gyal (whose name is dropped in the AI Bad Bunny lyrics) posted a video on TikTok, dancing to that specific part, saying that she first thought it was Bad Bunny and was disappointed to find out it was AI.

Bad Bunny was also disappointed. The song's incredible success on social media, which had people comment that might be even better than the original Bad Bunny songs, led the singer to address his fans saying that "anyone who likes this *insert inappropriate word here* is no longer my friend".

FlowGpt on the other hand, and creator of “NostalgIA'' using AI, came back with an answer, offering peace, and pointing out the beauty of creativity that new technologies can bring. FlowGpt even made an offer for a collaboration to Bad Bunny, preaching “Unity, peace, love, and a lot of reggaeton”.

It is not the first time that the voice of a famous singer is used without their knowledge or consent to create an AI generated hit.

Earlier this year, the song “Heart on my Sleeve” featuring AI-generated renditions of vocals by Drake and The Weeknd left a great impression to the public, and even led to rumors of it being nominated for a Grammy. Frank Ocean's vocals have also been artificially worked into songs that he did not even record, scamming a lot of his fans, who were hungry for new music and paid for it, to then realize it was just AI.

AI-generated music has also taken music platforms such as Spotify by storm. The platform is overflowing with songs made by AI and only this year deleted tens of thousands of them, trying to fight against ‘fraud’. In any event, it is worth noting that the decision to delete did not stem from the artificial origin of the music, but rather from the fact that the songs were being used to collect royalties on behalf of fraudulent accounts. While AI is not banned in all forms on the platform, the company does not allow its content to be used to train a machine learning or AI model, which can then produce music.

But is AI a threat to music, and to artists' voice, style, and personality?

Reggaeton haters have claimed that it is the simplicity of this genre that makes reggaeton artists' style and way of making lyrics, or even imitating the way they command their voice, so easy to recreate through AI. However, AI technology is getting so good that even an amateur can create a song worth listening to, in any style of music or artist, and do so fast.

Undoubtedly, there are very creative people like the ones behind FlowGpt, that do not have access to the music industry as we know it and that are getting a say through AI. However, putting ourselves in the shoes of the artists whose voice and art are being used to feed a machine that will then copy everything that makes them who they are, it is not difficult to comprehend Bad Bunny's reaction.

Did the track infringe Bad Bunny's rights? What other avenues do the artists have to fight AI-generated music?

The debate of who owns AI-generated music is ongoing. In the US for example, material that has been created by AI but has not been “touched” by a human author cannot be copyrighted, while the Beijing Internet Court recently ruled that AI-generated images can be protected by copyright at certain conditions. In the case of Bad Bunny, what rights do human owners have when AI creates something in the style of someone else? Keeping in mind how AI worked in this case, we would have to look at the copyright issues involved with the music that was fed to the machine (the “input”) and the outcome of the machine's “work” (the “output”).

Concerning input, one of the ways in which AI can learn to mimic voices of singers or compositional styles is by being trained on large amounts of data, known as "text and data mining" (TDM). FlowGPT would have to train the AI by having it listen to a whole bunch of Bad Bunny songs (fun!), probably without having Bad Bunny´s consent. Essentially, FlowGPT would have fed the copied songs into the system, and that could mean copyright infringement.

However, the argument might be that AI training is not something that’s being done to make a copy for commercial purposes; instead, it is deliberately transformative and trying to create something new. As such it could be argued that it has no direct impact on the market for the original by allowing an AI to hear or see a copy to be able to create by training its algorithm. Currently, TDM exceptions for non-commercial use exist in some jurisdictions such as the UK (Article 29A, CDPA), whereas in the EU the exception in Article 4 of the DSM Directive applies to any type of use, but rightsholders have the ability to opt their works out of the TDM exception. In the US such actions might be allowed based on the “fair use” doctrine due to the transformative character of the issue at hand.

Concerning the lawfulness of the outcome of this machine training, we need clarity on the exclusivity of making derivative works. If copyright law gives exclusive rights to create derivatives based on prior work to the owner, does using AI to create something based on that work make it a derivative only the original owner can produce?

Making music mimics the style of someone else can be allowed, for example if it falls under the pastiche exception. In this case, the part of Bad Bunny in the song could be argued to be only a small one, and the style attributable to him could be narrowed down to the way that some words are pronounced. But where we have machine learning and AI generating the work, it's an open question as to whether those outputs themselves are protected. In case an argument of Bad Bunny was that “NostalgIA” is a copy or a derivative of one of his songs, FlowGPT could answer that the lyrics are different, the music is different, therefore it´s a different song, and no rights of Bad Bunny exist in that song. If a person were to listen to a lot of Bad Bunny music and write their own song, using him and his style as an inspiration only, without using the music or lyrics of that song, it could be hard to prove that it is a copy. Why wouldn’t we use the same standard for an AI?

But mimicking someone's voice?

Even if the music and lyrics can be considered as something new, wouldn't a voice that confuses even hard-core fans be considered as the same voice? In some jurisdictions, including a number of US states, musical impersonation of a famous musician is considered to infringe the musician´s right of publicity, which prevents the unauthorized commercial use of an individual's likeness, or other recognizable aspects of one's persona, such as their voice. It gives an individual the exclusive right to license the use of their identity for commercial promotion. Such was the case of Bette Midler against Ford in 1988, when Ford produced a commercial and since Bette Midler refused to give permission to that, they hired an impersonator to sing the song. The Court held that when you have a distinctive voice of a professional singer that is intentionally imitated to sell something, that’s a violation of the right of publicity under state law.

In sum

The time of AI music is upon us - whether for the good or the bad no one can yet tell. On the one hand, our creativity possibilities become endless. Anyone could turn a murmur of their voice into a symphony, and write a whole song with the help of AI in seconds just by giving a simple prompt. On the other hand, anyone could produce a song, without being a singer, writer or producer at all, just by clicking a button, and feeding an algorithm, copying the style and voice of others, with very minimal implication of personal creativity. And AI algorithms become better every day. Will our copyright laws be ready to keep up and protect the humans that made music in the first place?