On 2 November 2023, a dead man released a new song. John Lennon, murdered outside his Manhattan apartment building in December 1980, sang lead vocals on “Now and Then,” the final Beatles single, almost 43 years after his killing. His voice was not synthesised, not cloned, not approximated by an algorithm trained on his catalogue. It was his actual voice, recorded on a cheap cassette player at the Dakota building sometime around 1977, rescued from decades of technical oblivion by machine learning software that could do what no human engineer had managed in nearly three decades of trying: separate his singing from the piano bleeding through beneath it.

The technology that made this possible, a neural network called MAL (a double homage to the HAL computer in 2001: A Space Odyssey and the Beatles' road manager Mal Evans), was developed by Peter Jackson's WingNut Films during the production of the documentary series Get Back. Its purpose was straightforward if technically extraordinary. MAL could be taught to recognise individual sound sources within a mono recording and then isolate them, pulling apart instruments and voices that had been fused together on a single track. As Giles Martin, the song's co-producer and son of legendary Beatles producer George Martin, explained to Variety: “Essentially, what the machine learning does is it recognises someone's voice. So if you and I have a conversation and we're in a crowded room and there's a piano playing in the background, we can teach the AI what the sound of your voice, the sound of my voice, and it can extract those voices.”

That technical feat unlocked something that had been attempted and abandoned twice before. It also raised a question that reverberates far beyond a single pop song, however beloved: when artificial intelligence enables the completion of an artist's unfinished work decades after their death, what kind of creative act is that, exactly? And once the precedent has been set, with a Grammy Award as validation, who gets to decide which ghosts sing next?

A Cassette Labelled “For Paul”

The story of “Now and Then” begins with grief and a cassette tape. In January 1994, Paul McCartney approached Yoko Ono, believing she might have some of Lennon's unused recordings. Ono gave McCartney three cassettes from Lennon's so-called retirement period in the late 1970s, when he had stepped back from public life to raise his son Sean at the Dakota. One cassette bore the words “For Paul” in Lennon's own handwriting. It contained rough piano-and-vocal demos of four songs: “Free as a Bird,” “Real Love,” “Grow Old with Me,” and “Now and Then.”

The first two songs became reunion singles during the Beatles' Anthology project in 1995 and 1996, produced by Jeff Lynne of the Electric Light Orchestra. Both reached the charts. Both featured new instrumental contributions from McCartney, George Harrison, and Ringo Starr layered around Lennon's demos. “Now and Then” was supposed to be the third.

On 20 and 21 March 1995, the three surviving Beatles gathered in the studio to work on it. The session did not go well. A persistent 60-cycle mains hum saturated the recording. Lennon's voice and piano were locked together on the same track, meaning any attempt to raise the vocal also raised the piano. The noise reduction software available at the time, a Pro Tools plugin called DINR, could not adequately clean the tape. Jeff Lynne spent two weeks trying at his home studio. The results were unsatisfying. “It was one day, one afternoon, really, messing with it,” Lynne later explained. “The song had a chorus but is almost totally lacking in verses. We did the backing track, a rough go that we really didn't finish.”

There was also the matter of George Harrison's opinion. McCartney later recalled that Harrison had dismissed the song as “fucking rubbish,” though Harrison's widow, Olivia, offered a gentler interpretation before the song's eventual release. “Back in 1995, after several days in the studio working on the track, George felt the technical issues with the demo were insurmountable and concluded that it was not possible to finish the track to a high enough standard,” she said. “If he were here today, Dhani and I know he would have whole-heartedly joined Paul and Ringo in completing the recording of 'Now and Then.'”

Harrison died in November 2001. The song sat on a shelf for another two decades.

The Machine That Heard What Humans Could Not

The breakthrough arrived from an unexpected direction. During the production of Get Back, Peter Jackson's team confronted a similar audio problem at massive scale: 60 hours of footage from the Beatles' January 1969 recording sessions, much of it captured by a single microphone that had picked up instruments, voices, and ambient noise in an undifferentiated jumble. The documentary would have been impossible without a way to separate those sounds.

Jackson's team, working with dialogue editor Emile de la Rey and machine learning researcher Paris Smaragdis at the University of Illinois Urbana-Champaign, built MAL from scratch. They scoured academic papers on audio source separation, determined that existing research was insufficient for their purposes, and created their own training data at a quality level that surpassed what had been used in prior academic experiments. The neural network was fed isolated recordings of individual Beatles instruments and voices, learning the spectral signature of each until it could reliably distinguish John from Paul, guitar from bass, drums from background chatter.

As Jackson described the process: “We developed a machine learning system that we taught what a guitar sounds like, what a bass sounds like, what a voice sounds like. In fact we taught the computer what John sounds like and what Paul sounds like. So we can take these mono tracks and split up all the instruments.”

When McCartney saw what MAL could do for the documentary, the connection was immediate. If the software could untangle the sonic chaos of the Twickenham sessions, perhaps it could also rescue Lennon's vocal from that stubborn cassette. It could. Within seconds, according to McCartney, the machine stripped away the piano and the hum, leaving Lennon's voice isolated and clear. “They said this is the sound of John's voice,” McCartney recalled. “A few seconds later and there it was, John's voice, crystal clear. It was quite emotional.”

Giles Martin was emphatic about what had and had not happened. “AI is not creating John's voice,” he told MusicRadar. “John's voice existed on that cassette and we made the song around him.” The distinction matters enormously. No synthetic voice was generated. No words were invented. No performance was fabricated. The technology's role was purely subtractive: removing what obscured a real human performance so that it could finally be heard.

Building Around a Ghost

With Lennon's vocal isolated, the completion of “Now and Then” became a conventional, if emotionally charged, production exercise. McCartney recorded new bass, a slide guitar solo in the style of Harrison as a tribute, electric harpsichord, backing vocals, and piano that echoed the feel of Lennon's original demo. Starr laid down a finalised drum track and added backing vocals. Harrison's guitar parts, both acoustic and electric, recorded during the abandoned 1995 sessions, were extracted and incorporated.

Rather than use AI to recreate the Beatles' signature vocal harmonies, Martin took a more analogue approach. He pulled actual Beatles vocal recordings from existing multitrack tapes of songs like “Eleanor Rigby,” “Because,” and “Here, There and Everywhere,” and wove them into the arrangement. “I'm not using AI to recreate their voices in any way,” Martin told interviewers. “I'm literally taking the multitrack tapes.” He added, with characteristic directness: “It might have been easier if I used AI, but I didn't.”

A string arrangement written by McCartney, Martin, and Ben Foster was recorded at Capitol Studios. The result was a song that featured all four Beatles: Lennon's 1977 vocal, Harrison's 1995 guitar, and McCartney and Starr's 2022 contributions, a creative object spanning 45 years of performances by musicians who were never all in the same room for this particular song and two of whom were dead by the time it was finished.

Validation and the Weight of a Grammy

The commercial and institutional response was striking. “Now and Then” debuted on the UK Singles Chart on 3 November 2023 and reached number one the following week, becoming the Beatles' 18th UK number one and their first in 54 years, since “The Ballad of John and Yoko” in 1969. It set the record for the longest gap between number one singles by any musical act. At the ages of 81 and 83 respectively, McCartney and Starr became members of the oldest band to claim a UK number one single. The single was the fastest-selling vinyl release of the century in the UK, with 19,400 copies sold on vinyl alone, and accumulated 5.03 million streams in its first week, the most ever for a Beatles track.

Then came the Grammy. On 2 February 2025, “Now and Then” won Best Rock Performance at the 67th Annual Grammy Awards, beating out songs by Pearl Jam, IDLES, the Black Keys, St. Vincent, and Green Day. It was the Beatles' first Grammy win since 1997, when they had won for “Free as a Bird,” itself a posthumously completed Lennon demo. It was also, historically, the first AI-assisted track to win a Grammy Award.

Neither McCartney nor Starr attended the ceremony. Sean Ono Lennon, John's son with Yoko Ono, accepted the award. “Since no one is coming up to take this award, I figured I'd come and sit in,” he said. “I really didn't expect to be accepting this award on behalf of my father's group, the Beatles.”

The Grammy matters not merely as an honour but as a legitimising act. The Recording Academy, by bestowing its most prestigious recognition on a track that could not have existed without machine learning, effectively declared that this kind of creative act falls within the boundaries of what the music industry considers real, valid, and worthy of its highest prizes. That declaration will be difficult to walk back.

A New Category, or an Old Power Reasserted

Here is where the philosophical terrain gets uneven. The careful, collaborator-blessed, estate-approved process behind “Now and Then” can be read in two fundamentally different ways.

The first reading is optimistic, even utopian. This is a genuinely new kind of creative act, one that exists outside traditional notions of single authorship. No individual made this song. Lennon wrote the melody and sang the vocal but never finished the composition and could not consent to its completion. Harrison contributed guitar parts in 1995 for a song he openly disliked, and his participation in the final version was sanctioned by his widow and son rather than by the man himself. McCartney and Starr completed the arrangement nearly three decades after the aborted sessions, working with a producer (Giles Martin) who had not been involved in the original attempt. The technology that made it possible was developed for an entirely different project by a filmmaker from New Zealand. The result is a creative object with no single author, no unified moment of creation, and no clear boundary between human artistry and machine capability.

The second reading is more sceptical. Strip away the sentiment, and what happened is that the surviving members of a band, along with their associated estates and production teams, used new technology to finish a project on their terms, shaping how a dead colleague is remembered in a way that he cannot contest. Harrison called the song “fucking rubbish” in 1995. Lennon never heard a finished version of any kind. The decision to release “Now and Then” was made entirely by living people (McCartney, Starr, the Lennon estate, the Harrison estate) with commercial and emotional interests in the outcome. Olivia Harrison's statement that George “would have whole-heartedly joined” the project if he were alive is precisely the kind of claim that cannot be tested. It is an assertion of posthumous consent by someone who is not the deceased.

This is not to impugn anyone's motives. By every available account, the completion of “Now and Then” was undertaken with genuine love and reverence for the material, with painstaking care over the production, and with the blessing of all relevant estates. But the power dynamics are worth noting: it is always the living who decide how the dead are heard.

Precedent and the Catalogue of the Unfinished

The Beatles are not the first case of AI assisting in the completion of a deceased artist's unfinished work, but they are the most culturally significant. In October 2021, a team led by Professor Ahmed Elgammal of Rutgers University and Austrian composer Walter Werzowa premiered an AI-completed version of Beethoven's Tenth Symphony at the Telekom Forum in Bonn, Germany. The project had been organised by Matthias Roder, director of the Karajan Institute in Salzburg, to mark the 250th anniversary of Beethoven's birth. The AI was trained on Beethoven's complete body of work and the surviving sketches for the Tenth Symphony, generating hundreds of musical variations each day from which Werzowa selected the most plausible continuations. The result was two complete movements of more than 20 minutes each. When the team challenged an audience of experts to determine where Beethoven's phrases ended and where the AI extrapolation began, they could not.

AIVA, the Artificial Intelligence Virtual Artist, has similarly completed an unfinished Dvořák piano composition in E minor, and various projects have tackled Schubert's “Unfinished” Symphony. In each case, the technical achievement was impressive, but the cultural stakes were comparatively low. Classical music has a long tradition of scholarly completions; Deryck Cooke's performing version of Mahler's Tenth Symphony, for example, has been in concert repertoire since the 1960s. The idea that someone other than the original composer might finish an unfinished symphony is not alien to that world.

Popular music is different. The connection between artist and audience is more personal, more identity-driven, more commercially charged. When a rock or pop artist's unfinished recordings become candidates for technological resurrection, the questions multiply. Whose vault gets opened next? What constitutes sufficient source material for a legitimate completion? If the Beatles' approach represents the gold standard, with surviving collaborators overseeing the process, what happens when there are no surviving collaborators? What happens when the estate holders have financial incentives that may not align with artistic ones?

The music catalogue acquisition market offers a sobering context. According to MIDiA Research, the value of music catalogue acquisitions since 2010 has reached at least 6.5 billion dollars in publicly disclosed transactions alone. Prince's estate sold nearly 50 per cent of rights to his name, likeness, masters, and publishing to Primary Wave. Michael Jackson's estate cashed out his 50 per cent stake in Sony/ATV for 750 million dollars in 2016. When a catalogue is worth hundreds of millions, the financial pressure to generate new revenue from it is enormous. An AI-completed “new” track from a deceased superstar represents a potential commercial event of the first order.

The Dark Mirror of Unauthorised Resurrection

If “Now and Then” represents the careful, consensual end of the spectrum, the opposite extreme is already flourishing. In April 2024, during his feud with Kendrick Lamar, Drake released “Taylor Made Freestyle,” a track featuring AI-synthesised vocals of the late Tupac Shakur. The response from Tupac's estate was swift and furious. Howard King, the estate's attorney, sent a cease-and-desist letter calling Drake's use “a flagrant violation of Tupac's publicity and the estate's legal rights” and “a blatant abuse of the legacy of one of the greatest hip-hop artists of all time.” King added that “the Estate would never have given its approval for this use.” Drake removed the track within days. The irony was not lost on observers: Drake's own label had previously taken down “Heart on My Sleeve,” a 2023 track by an anonymous creator that used AI to clone the voices of Drake and the Weeknd without permission.

By 2025, the problem had moved far beyond individual celebrity disputes. An investigation by 404 Media found that AI-generated tracks were being uploaded to the official Spotify profiles of deceased musicians without any permission from their estates. Blaze Foley, a Texas folk singer who died in 1989, had a synthetic song called “Together” appear on his verified Spotify page, uploaded via TikTok's SoundOn distribution platform. Grammy-winning songwriter Guy Clark, who died in 2016, had an AI-generated song placed under his name. The electro-pop artist Sophie, who died in 2021, and Uncle Tupelo, the former band of Wilco's Jeff Tweedy, were similarly targeted.

The mechanism is disturbingly simple. Independent distribution services like DistroKid, TuneCore, and SoundOn serve as intermediaries between artists and streaming platforms. Spotify relies on these “trusted” distributors to provide accurate metadata but does not independently verify whether an artist is alive, whether the submitter has rights to the name, or whether the music is genuine. Anyone with access to AI music generation tools like Suno or Udio can create a plausible imitation of a real artist in seconds and upload it through these distribution channels. The fake track then appears alongside the artist's legitimate catalogue, indistinguishable to casual listeners.

Spotify has said it removed 75 million “spammy” tracks in a single year and launched a tool for artists to report mismatched releases. But the company has no system for tagging or labelling AI-generated music and has not disclosed how it identifies such content. The scale of the problem is significant: Deezer has reported that 18 per cent of all music uploaded to streaming platforms is fully AI-generated.

Legislative Scaffolding in Progress

The legal landscape is evolving rapidly, though it has not yet caught up with the technology. Tennessee's ELVIS Act (Ensuring Likeness, Voice, and Image Security Act), signed into law by Governor Bill Lee on 21 March 2024, was the first enacted legislation in the United States specifically designed to protect musicians from unauthorised AI voice cloning. The bill passed both chambers of the Tennessee legislature unanimously, reflecting the state's deep ties to its music industry, which supports more than 60,000 jobs and contributes 5.8 billion dollars to the national GDP.

The ELVIS Act grants individuals rights over their voice “regardless of whether the sound contains the actual voice or a simulation of the voice of the individual” and imposes liability on technology providers, not merely end users. It protects both living and deceased individuals from digital exploitation. California has pursued similar measures, updating its long-established right-of-publicity laws to explicitly cover AI-based infringements.

At the federal level, the No AI FRAUD Act would establish a national right in an individual's likeness and voice, while the NO FAKES Act would create liability for the production or distribution of unauthorised AI-generated digital replicas in audiovisual works or sound recordings. Neither had been enacted as of early 2026, leaving protection largely dependent on a patchwork of state laws.

These measures address the most egregious abuses: outright voice cloning, unauthorised deepfakes, fraudulent streaming uploads. What they do not address is the murkier territory that “Now and Then” occupies. When surviving collaborators and authorised estates use emerging technology to complete an unfinished work, existing legal frameworks generally permit the activity. The question is not legality but legitimacy, and that is a cultural judgement rather than a statutory one.

Commercial Gravity and the Erosion of Restraint

The commercial incentives pushing towards more AI-assisted posthumous completions are substantial and growing. Every major record label sits on vaults of unreleased material by deceased artists. Prince alone left behind an estimated 8,000 unreleased songs in his vault at Paisley Park at the time of his death in 2016, enough material, by some estimates, for his estate to release an album a year for a century. The potential to transform these recordings into finished, releasable tracks using the same techniques applied to “Now and Then” represents an enormous financial opportunity.

The restraint shown in the Beatles' case was enabled by several unusual factors. McCartney and Starr are independently wealthy and had no financial need to release the song. The Beatles' catalogue was already one of the most commercially successful in music history, meaning marginal revenue from one additional single was not a decisive factor. The surviving principals had genuine personal connections to the material and the deceased artists. And the public narrative, “the last Beatles song,” had a built-in emotional arc that encouraged care rather than exploitation.

Remove any of these factors and the calculus shifts. An estate managed by distant relatives or corporate entities, a catalogue whose value depends on generating new releases, a fanbase hungry for any scrap of unreleased material: these conditions are ripe for a less restrained approach. The technology that separated Lennon's voice from a cassette hum can just as easily be applied to bootleg recordings, rehearsal tapes, isolated vocal takes, and fragmentary demos by any artist whose voice can be used as training data for source separation algorithms.

The question is not whether this will happen but how quickly commercial pressure will override the curatorial care that characterised “Now and Then.” The Grammy win accelerates this timeline. When the music industry's most prestigious institution rewards an AI-assisted posthumous completion, it sends an unmistakable signal to every label, estate, and producer with access to a deceased artist's unreleased recordings: this is not merely acceptable, it is excellent. It wins awards. It reaches number one.

The Living and the Dead

There is a deeper discomfort at work, one that transcends the specifics of the Beatles or any individual artist. The history of posthumous releases is littered with cautionary tales. After Michael Jackson's death in 2009, his estate released the album Michael in 2010, which sparked fierce controversy when Jackson's own family members claimed that three tracks featured vocals by an impersonator rather than by Jackson himself. After more than a decade of fan protest and legal action, the disputed songs were eventually removed from streaming platforms. His estate later released Xscape in 2014, taking greater care to preserve Jackson's authentic vocal performances, but the earlier debacle had already demonstrated how readily commercial interests could override questions of authenticity. After Prince's death in 2016, the management of his vault became a matter of intense legal and familial dispute, with his estate passing through intestacy laws in the absence of a will.

AI does not create these tensions. It amplifies them. When the technological barrier to finishing an unfinished song drops to near zero, the only remaining barriers are ethical, legal, and cultural. And history suggests that ethical and cultural barriers erode faster than legal ones when significant money is at stake.

Paul McCartney himself framed his decision in terms of imagined consent. “Is it something we shouldn't do?” he told interviewers. “Every time I thought, like that, I thought, 'wait a minute. Let's say I had a chance to ask John. Hey John, would you like us to finish this last song of yours?' I'm telling you, I know the answer would have been 'yeah.'”

McCartney may well be right. But the logic of imagined consent is infinitely extensible. Anyone close to a deceased artist can claim to know what that artist would have wanted. The closer the relationship, the more credible the claim, but it remains fundamentally untestable. And as the distance between the deceased artist and the people making decisions about their legacy grows, from bandmates to widows to children to grandchildren to corporate entities holding catalogue rights, the claim of imagined consent becomes progressively thinner.

What Comes After the Last Beatles Song

“Now and Then” is a beautiful, melancholy record. It sounds like the Beatles, because in every meaningful sense it is the Beatles. Lennon's voice is his own. Harrison's guitar is his own. McCartney and Starr played their parts with the skill and sensitivity of men who spent their formative years making music together. The machine learning software that made it possible did not create anything; it revealed what was already there but hidden.

And yet the song exists because living people decided it should, using capabilities that did not exist when the dead had any say in the matter. That is the irreducible fact at the centre of this story, and it will only become more significant as the technology improves, as the vaults open wider, and as the commercial logic of the music industry seeks new revenue from old recordings.

So is this a fundamentally new category of creative act? In one sense, yes. No previous generation of musicians had access to tools that could extract a voice from a degraded cassette with such fidelity, making collaboration across decades and beyond death a technical reality rather than a metaphor. But in another sense, the answer is less comforting. The power to decide what the dead would have wanted has always belonged to the living. AI does not redistribute that power; it supercharges it. The careful restraint of the Beatles' approach deserves recognition and respect. It also deserves to be understood for what it is: a best-case scenario, executed by people with the resources, the relationships, and the cultural authority to do it well. The next case may not look like this. The case after that almost certainly will not. The technology that gave us one last Beatles song will not stop there. The question is whether the industry, the legal system, and the culture can build frameworks of care and consent that match the capabilities of the machines. On current evidence, the machines are moving faster.

---

References and Sources

1. “Now and Then (Beatles song).” Wikipedia. https://en.wikipedia.org/wiki/Now_and_Then_(Beatles_song

2. Fortune Europe. “Paul McCartney, Ringo Starr and Peter Jackson used AI for 'separating' a John Lennon vocal to make the very last Beatles song ever.” October 2023. https://fortune.com/europe/2023/10/26/last-beatles-song-using-ai-now-and-then-peter-jackson-paul-mccartney-john-lennon/

3. NPR. “How producers used AI to finish The Beatles' 'last' song, 'Now And Then.'” November 2023. https://www.npr.org/sections/world-cafe/2023/11/02/1208848690/the-beatles-last-song-now-and-then

4. Rolling Stone. “The Beatles Return for One More Masterpiece With New Song 'Now and Then.'” November 2023. https://www.rollingstone.com/music/music-news/beatles-new-song-now-and-then-1234868643/

5. The Conversation. “Now and Then: enabled by AI, created by profound connections between the four Beatles.” November 2023. https://theconversation.com/now-and-then-enabled-by-ai-created-by-profound-connections-between-the-four-beatles-216920

6. MusicRadar. “Giles Martin explains why you'd be wrong to think 'AI' created Lennon's parts for The Beatles' Now and Then.” https://www.musicradar.com/artists/giles-martin-ai-beatles-now-and-then

7. Variety. “Giles Martin on Producing the Beatles' 'Now and Then,' Remixing the Red and Blue Albums.” November 2023. https://variety.com/2023/music/news/beatles-giles-martin-now-and-then-producer-remixing-red-blue-albums-interview-1235778746/

8. MusicTech. “It might have been easier if I used AI, but I didn't: How Giles Martin created the backing vocals for The Beatles' Now and Then.” https://musictech.com/news/music/giles-martin-beatles-now-and-then-production-ai/

9. Official Charts. “The Beatles' Now And Then is UK's Official Number 1 song in record-breaking return.” November 2023. https://www.officialcharts.com/chart-news/beatles-now-then-number-1-song-record/

10. CNN. “The Beatles break UK chart records as 'Now and Then' becomes No. 1 single.” November 2023. https://www.cnn.com/2023/11/11/entertainment/the-beatles-break-uk-chart-records-as-now-and-then-becomes-no-1-single/index.html

11. The Beatles Official Website. “Now And Then wins GRAMMY for Best Rock Performance.” February 2025. https://www.thebeatles.com/now-and-then-wins-grammy-best-rock-performance

12. Consequence of Sound. “The Beatles' 'Now And Then' wins Best Rock Performance at 2025 Grammys.” February 2025. https://consequence.net/2025/02/the-beatles-win-best-rock-performance-2025-grammys/

13. Loudwire. “The Beatles Make History With First of Its Kind Win at 2025 Grammys.” February 2025. https://loudwire.com/beatles-history-first-of-its-kind-win-2025-grammys/

14. Smithsonian Magazine. “How Artificial Intelligence Completed Beethoven's Unfinished Tenth Symphony.” 2021. https://www.smithsonianmag.com/innovation/how-artificial-intelligence-completed-beethovens-unfinished-10th-symphony-180978753/

15. The Conversation. “How a team of musicologists and computer scientists completed Beethoven's unfinished 10th Symphony.” October 2021. https://theconversation.com/how-a-team-of-musicologists-and-computer-scientists-completed-beethovens-unfinished-10th-symphony-168160

16. NPR. “Team uses AI to complete Beethoven's unfinished masterpiece.” October 2021. https://www.npr.org/2021/10/02/1042742330/team-uses-ai-to-complete-beethovens-unfinished-masterpiece

17. Rolling Stone. “Tupac Estate Demands Drake Remove Taylor Made Freestyle Over AI Voice.” April 2024. https://www.rollingstone.com/music/music-news/tupac-estate-drake-remove-taylor-made-freestyle-ai-voice-1235009865/

18. NBC News. “Drake pulls 'Taylor Made Freestyle' after Tupac estate threatens action for apparent use of AI voice.” April 2024. https://www.nbcnews.com/pop-culture/pop-culture-news/drake-pulls-taylor-made-freestyle-tupac-estate-threatens-action-appare-rcna149592

19. Billboard. “Tupac Shakur's Estate Threatens to Sue Drake Over Diss Track Featuring AI-Generated Tupac Voice.” April 2024. https://www.billboard.com/pro/tupac-shakur-estate-drake-diss-track-ai-generated-voice/

20. 404 Media. “Spotify Publishes AI-Generated Songs From Dead Artists Without Permission.” July 2025. https://www.404media.co/spotify-publishes-ai-generated-songs-from-dead-artists-without-permission/

21. NPR. “When your favorite band's new song is an AI fake.” October 2025. https://www.npr.org/2025/10/27/nx-s1-5587852/spotify-ai-music-fakes

22. MusicTech. “Spotify posting AI-generated songs of dead artists without permission, new report reveals.” 2025. https://musictech.com/news/music/spotify-ai-generated-songs-dead-artists/

23. Wikipedia. “ELVIS Act.” https://en.wikipedia.org/wiki/ELVIS_Act

24. Latham & Watkins. “The ELVIS Act: Tennessee Shakes Up Its Right of Publicity Law and Takes On Generative AI.” 2024. https://www.lw.com/admin/upload/SiteAttachments/The-ELVIS-Act-Tennessee-Shakes-Up-Its-Right-of-Publicity-Law-and-Takes-On-Generative-AI.pdf

25. CNBC. “Paul McCartney says A.I. got John Lennon's voice on 'last Beatles record.'” June 2023. https://www.cnbc.com/2023/06/13/paul-mccartney-says-ai-got-john-lennons-voice-on-last-beatles-record.html

26. TechCrunch. “Don't be afraid of the 'AI-assisted' Beatles song, 'Now And Then.'” November 2023. https://techcrunch.com/2023/11/02/dont-be-afraid-of-the-ai-assisted-beatles-song-now-and-then/

27. MusicRadar. “Peter Jackson says that he used machine learning to restore the Beatles' music for Get Back documentary.” https://www.musicradar.com/news/the-beatles-audio-stems-get-back

28. The Beatles Bible. “Now And Then, song facts, recording info and more.” https://www.beatlesbible.com/songs/now-and-then/

29. Music Business Research. “AI in the Music Industry, Part 9: Finishing the Unfinished.” April 2024. https://musicbusinessresearch.wordpress.com/2024/04/01/ai-in-the-music-industry-part-9-finishing-the-unfinished/

30. Music Tech Policy. “Fake Tracks Are Exploiting Deceased Artists. The FTC Must Act.” August 2025. https://musictechpolicy.com/2025/08/01/fake-tracks-are-exploiting-deceased-artists-the-ftc-must-act/

31. CBS News/60 Minutes. “Exploring the unreleased music in Prince's vault.” April 2021. https://www.cbsnews.com/news/prince-welcome-2-america-60-minutes-2021-04-11/

32. Smooth Radio. “Inside Prince's vault where thousands of unreleased songs are reportedly still hidden.” https://www.smoothradio.com/artists/prince/unreleased-songs-music-vault/

33. TIME. “Why Drake Had to Remove A Song That Featured AI-Tupac Vocals.” 2024. https://time.com/6971720/drake-tupac-ai/

---

Tim Green
UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0009-0002-0156-9795
Email: tim@smarterarticles.co.uk