AI in the Music Industry – Part 19: The Ambivalent Relationship Between Music Major and AI Companies

As we saw in part 14 of the series using the example of the AI company Anthropic, the music majors are already taking legal action against copyright infringements that they believe are taking place through the training and use of AI models. At the same time, however, they are also supporting projects such as the AI-generated completion of the Beatles’ song “Now and Then” and are looking to collaborate with technology companies to keep up with the latest developments around AI in music. In this concluding part of the ‘AI in the music industry’ series, we take a closer look at the majors’ ambivalent relationship with the use of artificial intelligence in the creation and distribution of music.

A study by market research company Goldmedia, commissioned by the German GEMA and the French SACEM, estimates the market volume for generative music and audio AI at US $300 million for the year 2023. This corresponds to a market share of around 8 per cent of the market for generative AI applications. However, the study also highlights the massive growth potential of this market segment, which is forecast to reach a market volume of US $3.1 billion by 2028, with an average annual growth rate of around 60 per cent.[1]

The players in the music industry, above all the music majors, cannot ignore such economic potential and have already recognised the possibilities of AI in general and voice cloning in particular. Therefore, in August 2023, Costa Rican musician Pedro Capmany released a song called “Volveré” with support of the Warner Music Group’s Alternative Distribution Alliance (ADA), featuring his father José, who died in 2001 and is considered the founder of Costa Rican rock music.[2] The AI was trained with José Capmany’s recordings and his voice was recreated down to the smallest detail to immortalise it for the duet with his son. Warner boss Robert Kyncl, whose label controls the rights to José Capmany’s masters, was enthusiastic about the experiment and proudly announced that “Volveré” is “one of the first official and professionally AI-generated song featuring a deceased artist”.[3] This is only partly true, as Korean AI company Supertone has already recreated the voice of the deceased pop star Kim Kwang-seok in 2021.[4] In November 2022, Chinese company Tencent Music Entertainment was able to posthumously synthesise the voices of two long-dead Chinese superstars, Teresa Teng and Anita Mui, using AI technology known as Lingyin Engine.[5]

The Beatles’ song, ‘Now and Then’, which was completed by an AI, also made headlines. In June 2023, Paul McCartney announced in an interview on BBC 4 that he and Ringo Starr wanted to use an AI to complete the song, which John Lennon had recorded on cassette in his Manhattan apartment shortly before his death in 1978.[6] McCartney had received the tape from Lennon’s widow, Yoko Ono, in 1994 and the song should be included on ‘The Beatles Anthology’. At the time, however, it was impossible to separate Lennon’s voice from the piano accompaniment. The project gained new momentum after New Zealand star director Peter Jackson used AI to separate the Beatles’ voices from distracting background noise for a Beatles documentary in 2021. McCartney approached Jackson about using the same process on ‘Now and Then’, and the result was astonishing. Lennon and McCartney sing together one last time – thanks to artistic intelligence – accompanied by Ringo Starr and George Harrison, who had recorded his part for the anthology during his lifetime, as can be heard and seen in the music video produced by Peter Jackson.[7]

As well as being a cultural and historical milestone, ‘Now and Then’ was also good business for Universal. Within a week, the song stormed to number 1 in the official UK singles chart, giving the Beatles another record. In 1969, 54 years after “Come Together”, they made it back to number one. They even overtook the Rolling Stones as record holders, whose #1 songs “Living In A Ghost Town” and “Jumpin’ Jack Flash” were separated by 52 years.[8] Unsurprisingly, the CEO of Europe Central Europe, Frank Briegmann, immediately sent his congratulations: “The fact that ‘Now And Then’ reached number one in the German charts so many years later tells the story of this legendary band in a very special way. Congratulations to the Beatles and all the teams involved – it really is something very special [translation by the author]”.[9]

The majors are trying to avoid the mistakes they made in the past, when file-sharing emerged and they became mere passengers in a development they could no longer control. They are therefore looking to collaborate with AI companies and are also investing millions in AI applications. In May 2023, Universal Music Group signed an agreement with AI startup Endel, known for creating soundscapes, to help Universal artists create their own music projects using Endel’s AI.[10] In October 2023, Universal also announced a strategic alliance with Singapore-based BandLab Technologies. Among other things, the AI company operates the BandLab social media platform, which had 60 million registered users at the time of the deal. With the BandLab app, musicians can use AI to shape their creative ideas rhythmically, melodically and compositionally, and share them with other app users. The exact nature of the cooperation between the music major and the AI company is not clear from the press release[11] or the media coverage. [12]

Of particular note is the collaboration with Google subsidiary YouTube, which not so long ago was seen as a godsend for the music industry. Now, on 21 August 2023, the CEOs of YouTube and Universal released a statement announcing their intention to join forces and create the “Music AI Incubator” to experiment with AI-generated music. The consortium includes artists such as Brazilian singer Anitta, One Republic frontman Ryan Tedder, ABBA’s Björn Ulvaeus, as well as the Frank Sinatra estate and singer/songwriter Roseann Cash, heiress and eldest daughter of Johnny Cash.[13] In addition, YouTube CEO Neal Mohan has invited other music industry players to join the project and has published the three key principles that will guide the collaboration between AI developers and music industry players on his blog:

“Principle #1: AI is here, and we will embrace it responsibly together with our music partners.

Principle #2: AI is ushering in a new age of creative expression, but it must include appropriate protections and unlock opportunities for music partners who decide to participate.

Principle #3: We’ve built an industry-leading trust and safety organization and content policies. We will scale those to meet the challenges of AI.”[14]

At the same time, the Financial Times reported that YouTube’s parent company, Google, and Universal Music Group were in negotiations to licence vocal performances and melodies for use in AI systems.[15] In this case, the licence agreement would go far beyond the usual scope, as not only the use of music catalogues would be licensed, but also individual vocal performances and melodies which are not protected by copyright.

These negotiations have apparently led to a result, because on 16 November 2023, YouTube launched the “Dream Tracks” app as part of its YouTube Shorts, which allows users to generate voice clones of a variety of stars such as Charlie Puth, John Legend and Sia. This “experiment”, as YouTube calls it, is based on Google’s Deep Mind technology and is supported by both Universal and Warner.[16] On the same day as the app’s launch, Warner CEO Robert Kyncel, formerly CEO of YouTube, confirmed and welcomed the support of “Dream Tracks” to investors during the earnings call for Warner’s 2023 annual report.[17] But also the CEO of Universal. Lucian Grainge, welcomed YouTube’s AI experiment, saying: “(…) we must help artists achieve their greatest creative and commercial potential – in part by providing them access to the kind of opportunities and cutting-edge creative tools made possible by AI.”[18]

In the same statement, Grainge also emphasised that Universal Music Group has a fundamental responsibility to its artists to protect them and their work from unauthorised exploitation, including by generative AI.[19] This statement reflects the ambivalent attitude of music majors towards generative music AI. On the one hand, they are seeking to collaborate with technology companies such as Google, while on the other they are taking AI companies to court, as in the case of Anthropic, which is also funded by Google.

This ambivalence is also expressed in the Universal Music Group’s (UMG) statement to the US Copyright Office.[20] Universal emphasised that “(…) the music industry largely operates on a licensing model”[21] which sums up the dominant business model in the recording industry. The exploitation of music rights and music catalogues must be protected under all circumstances. UMG is certainly willing to license its copyrighted content to AI companies, similar to other uses such as music streaming: “UMG is ready and willing to explore licensing content to generative AI companies and has already established our disposition to work with those companies to find ways to build fair, lawful, and ethical business models that make economic, strategic, and long-term sense for our artists and songwriters.”[22] However, this requires direct licensing agreements between the music majors and the AI companies, and therefore the UMG rejects any form of collective licensing and compulsory licensing.[23] In its statement, Universal therefore argues in favour of an opting-in rule for training AI with musical content, rather than the opting-out solution as preferred by the technology companies.”An opt out system is based on the erroneous premise that training on copyrighted works without permission is by default lawful unless each copyright owner objects.”[24]

In any event, Universal does not consider the training of AI to be a ‘fair use’ practice, as it attempts to demonstrate in detail in its submission with precedents from US case law.[25] The music major thus takes the diametrically opposed position to Google, which sent its statement to the US Copyright Office at the end of October 2023 and formulated the following legal position: “The doctrine of fair use provides that copying for a new and different purpose is permitted without authorization where – as with training Al systems – the secondary use is transformative and does not substitute for the copyrighted work.”[26] According to Google, training AI is about acquiring knowledge in a similar way to reading a book or learning something. “The mere fact that, as a technological matter, copies need to be made to extract those ideas and facts from copyrighted works should not alter that result”,[27] Google points out that the reproduction process is covered by the ‘fair use’ doctrine. Currently, it is impossible to predict which position will prevail. It will probably take intensive negotiations between the music industry players and the technology companies to reach a compromise, otherwise the courts will have to decide.

It is noticeable that Sony is not a cooperation partner of YouTube. This could be because the company is pursuing its own AI strategy. At the end of November, Sony Music launched the website metallicspheres.io, which allows fans to use the new album “Metallic Spheres In Colour” by ex-Pink Floyd guitarists David Gilmour and The Orb as a starting point for new remixes with the help of generative AI. As well as Sony Music, the project is being supported by Vermillio, an AI company founded by Dan Neely in June 2021. Although the project is an experiment, Sony has stated that the concept will be realised with other Sony artists.[28] Overall, Sony is also ambivalent about generative AI music applications. Speaking at the US Senate’s AI Insight Forum on 28 November, Dennis Kooker, President of Global Digital Business & Sales at Sony Music Entertainment, emphasised that the company is looking to collaborate with AI companies and is currently in talks with 200 providers, but also called on policymakers to take legal action against the use of copyrighted music. His catalogue of demands to policymakers consists of five points: (1) When new music products are created with AI, the consent of rights holders must be obtained and they must be remunerated. (2) Ensure that the use of copyrighted music in AI training is not considered ‘fair use’. (3) The creation of voice clones must be prohibited unless prior consent has been obtained. (4) When training AI, it must be made transparent what content has been used so that ‘fair use’ accounting can be carried out. (5) AI-generated music must be labelled so that it can be clearly distinguished from human-generated music.[29] It can be assumed that Kooker’s demands echo those of his colleagues at other music majors and many indie labels. However, the demands also show that the record industry has been taken by surprise by the rapid development of AI in 2023.

In summary, all the music majors are looking at AI applications and supporting offerings. However, the impression is that ‘Fake Drake’ has triggered a flurry of activity among the majors to get to grips with the challenges of AI-generated music. This does not exactly look like a proactive AI strategy, but rather a reaction to developments that can no longer be ignored. The majors could have started adopting AI music applications for their own purposes a few years ago and taken the lead in driving development. Now it is the technology companies that are revolutionising music creation with new innovative AI technologies. However, 2023 will go down in history as the year AI was recognised as a game-changer for the music industry, ushering in a new era.

---

Endnotes

[1] Goldmedia, 2024, AI and Music. Market Development of AI in the Music Sector and Impact on Music Authors and Creators in Germany and France. A Report Commissioned by SACEM and GEMA, January 31, 2024.

[2] Music Business Worldwide, “AI a threat to music? Try telling that to this artist from Costa Rica – who’s just recreated his late father’s voice on a new track”, August 16, 2023, accessed: 2024-06-10.

[3] Ibid.

[4] Ibid.

[5] Ibid.

[6] Heise.de, “The Beatles: Ein allerletzter Song – mit Hilfe von KI”, June 13, 2023, accessed: 2024-06-10.

[7] YouTube, “The Beatles – Now and Then”, November 3, 2023, accessed: 2024-06-10.

[8] Musikexpress, “‘Now And Then’: Die Beatles nach 54 Jahren wieder auf Platz 1 der Charts” November 10, 2023, accessed: 2024-06-10.

[9] Ibid.

[10] Pitchfork, “Universal Music Group Signs Deal With AI Startup Endel”, May 24, 2023, accessed: 2024-06-10.

[11] Universal Music Group press release, “Universal Music Group and BandLab Technologies Announce First-of-its-Kind Strategic AI Collaboration”, October 18, 2023, accessed: 2024-06-10.

[12] Variety, “Universal Music and Bandlab Unveil AI Partnership Intended to Protect Artist Rights”, October 18, 2023, accessed: 2024-06-10.

[13] Music Business Worldwide, “YouTube and Universal Music Group partner to develop AI music tools – complete with ‘protections’ for artists and rightsholders”, August 21, 2023, accessed: 2024-06-10.

[14] Neal Mohan, “Our principles for partnering with the music industry on AI technology”, YouTube Official Blog, August 21, 2023, accessed: 2024-06-10.

[15] Financial Times, “Google and Universal Music negotiate deal over AI ‘deepfakes'”, August 8, 2023, accessed: 2024-06-10.

[16] Music Business Worldwide, “Welcome to the future: AI-generated vocal clones of superstars now available on YouTube Shorts – thanks to Google’s ‘most advanced music generation model to date'”, November 16, 2023, accessed: 2024-06-10.

[17] Music Business Worldwide, “WMG’s Robert Kyncl expects more artist-centric streaming deals, and 2 other things we learned on the company’s latest earnings call”, November 16, 2023, accessed: 2024-06-10.

[18] Music Business Worldwide, “Sony Music’s artists aren’t involved in YouTube’s new voice-cloning AI experiment. Not unrelated: Google’s recent filing with the US Copyright Office”, November 23, 2023, accessed: 2024-06-10.

[19] Ibid.

[20] Universal Music Group, 2023, Notice of Inquiry “Artificial Intelligence and Copyright” before the United States Copyright Office, Docket No. 2023-6.

[21] Ibid., p 39.

[22] Ibid., p 65.

[23] Ibid., p 66.

[24] Ibid., p 57.

[25] Ibid., pp 40-46.

[26] Google, 2023, “Artificial Intelligence and Copyright”, 88 Fed. Reg. 59942, Docket No. COLC-2023-0006, October 30, 2023, pp 1-2.

[27] Ibid., p 9.

[28] Music Business Worldwide, “Sony Music debuts generative AI music project with Pink Floyd’s David Gilmour and The Orb”, November 21, 2023, accessed: 2024-06-10.

[29] Music Business Worldwide, “Sony Music has issued nearly 10,000 deepfake takedowns … and other things we learned from Dennis Kooker’s speech about AI”, November 30, 2023, accessed: 2024-06-10.