Paul McCartney is working with archive recordings of John Lennon to finish a “last Beatles record” through AI, set to be released this year. Let’s dive into that.
What does that mean?
McCartney told BBC that Peter Jackson (director of the “Lord of the Rings” films and the 2021 “The Beatles: Get Back” documentary) was able to “extricate” Lennon’s voice “from a ropey little bit of cassette.” In this case, that means that some version of an “AI-powered” stem separation tool (think Audioshake, LALAL.AI, melody.ml) was applied to old Beatles demo tapes to pull out Lennon’s voice and generate stems that can be worked with further.
What doesn’t that mean?
Unlike the recent wave of “AI soundalikes” (think Ghostwriter977’s viral “Heart on My Sleeve” using the AI-synthesized voices of Drake and The Weeknd), the AI tech at play here is not recreating or synthesizing Lennon’s voice from a set of training data. AI soundalikes are created using a diffusion model, a type of generative AI model trained on the desired voice, and will only work if the input is a recording of a voice. Stem splitting, as Peter Jackson used here, takes a recording with vocals, instrumentals, and any other sounds together as the input in order to output separate tracks of its constituent parts.
What are the legal implications of AI in music?
Taking a step back from The Beatles, there are numerous ongoing arcs in the AI/music/law story. Namely: the use of copyrighted works in training datasets, artists’ right to own their voices, record labels grasping for control over artists’ voices, and more.
Is it fair use to train AI on copyrighted works? One former United States Copyright Office General Counsel, Sy Damle, testified before Congress last month that it is. Days later, another Copyright Office General Counsel, Jon Baumgarten, emphatically pushed back in a letter. With such diametrically opposed viewpoints from different former General Counsel of the Copyright Office, it seems clear that shifts in future leadership will cause drastic changes to the underlying policy surrounding this issue. The current Copyright Office leadership has not weighed in.
One area where the Copyright Office has provided guidance, however, deals with the protection of works created or influenced by AI. According to the Copyright Office, the general rule is that AI-generated works will not be eligible for copyright registration. They have stated, however, that “a work containing AI-generated material will also 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.’ (citing 17 U.S.C. 101 (definition of ‘compilation’)). . . . Or an artist may modify material originally generated by AI technology to such a degree that the modifications meet the standard for copyright protection. (Citing Compendium (Third) sec. 507.1 (identifying that where a new author modifies a preexisting work, the “new authorship . . . may be registered, provided that it contains a sufficient amount of original authorship”); see also 17 U.S.C. 101 (defining ‘derivative work’ to include works ‘based upon one or more preexisting works’ where modifications to the work ‘which, as a whole, represent an original work of authorship’)). 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. (citing 17 U.S.C. 103(b)).”
In this guidance, the Copyright Office has acknowledged that “[a]uthors have long used such tools to create their works or to recast, transform, or adapt their expressive authorship” and that AI tools can continue to aid artists in their creative process. To that end, the policy is that “what matters is the extent to which the human had creative control over the work’s expression and ‘actually formed’ the traditional elements of authorship.”
So, there’s some guidance there, albeit not incredibly clear yet. Namely, how much human input is required to make an AI-generated song copyrightable? The Copyright Office seems to have no desire to set a threshold or bright line by which artists (and their attorneys) can set expectations, so this remains an open area for future litigation and/or legislation.
To the extent that the upcoming Beatles record derived in part from AI technology, the Copyright Office would look to the nature and extent of the AI’s role in its creation in determining whether it is protectable by copyright. Given that AI was used solely as a tool to extract John Lennon’s vocal stems from existing demo recordings, and that the creative process behind the songwriting, musical composition, production, mixing, and mastering were all done in such a way where “human[s] had creative control over the work’s expression,” it is highly likely that the Copyright Office would deem the work suitable for copyright registration.
What should we be on the lookout for?
A new Beatles song was not on anybody’s 2023 bingo card. This will be the first high-profile example of a legendary act tapping into AI to enable a new sort of comeback, but I suspect it will not be the last. If the new Beatles release is well-received, I would be on the lookout for future releases from Michael Jackson, Whitney Houston, Elvis, and who knows who else over the coming years.
Reprinted with permission from the June 28, 2023 edition of The Legal Intelligencer © 2023 ALM Global Properties, LLC. All rights reserved. Further duplication without permission is prohibited, contact 877-256-2472 or asset-and-logo-licensing@alm.com.