The lawsuits filed by the majors against AI companies Suno and Udio could be the most important cases for the music industry since the Supreme Court's Grokster decision, as I explained in last week's Follow the Money column. The results are hard to predict, however, because the central issue will be “fair use,” a US legal doctrine shaped by court decisions that includes famous — sometimes infamous — nuances about art and appropriation. And while most creators are more focused on issues around productive AI “outputs” — music they'll have to compete with or songs that might sound similar to their own — those cases include the legality of copying music for AI training.
Neither Suno nor Udio have said how they train their AI programs, but both have essentially said that copying music to do so would qualify as fair use. The specification could touch on Google Books development, Android OS compatibility, and more Supreme Court case featuring Prince, Andy Warhol and Vanity fair. It's the kind of fair use case that once inspired a judge to call copyright “the metaphysics of law.” So let's go metaphysical!
Fair use essentially provides exceptions to copyright, usually for the purpose of free expression, allowing, among other things, reference (as in book or movie reviews) and parody (to comment on art). (The iconic example in music is the Supreme Court case over 2 Live Crew parody (Roy Orbison's “Oh, Pretty Woman.”) These determinations include a four-factor test that weighs “the purpose and character of the use.” “the nature of the copyrighted work”; how much and how significantly a part of the work is used. and the effect of the use on the potential market value of the copyrighted work. However, in the last decade or so, the concept of “transformative use”, derived from the first factor, has been expanded in a way that has allowed the development Google Books (the copying of books to create a database and excerpts) and its use some Oracle API code in Google's Android system — which we could say goes beyond the origin of the idea.
Could copying music for machine learning purposes also qualify?
In a paper on the subject, “Fair use in the US Redux: Reformed or even distorted,” the distinguished professor at Columbia Law School Jane Ginsburg suggests that the influence of the transformative use argument may have reached its peak. (I'm oversimplifying a very smart paper, and if you're interested in this topic, you should read it.)
The Supreme Court's decision in the Google-Oracle case involved part of a computer program, away from the creative “core” of copyright, and musical recordings would probably be judged differently. The Supreme Court also made a very different decision last year in a case brought by the Andy Warhol Foundation for the Visual Arts against a prominent rock photographer Lynn Goldsmith. The case involved an Andy Warhol screen print of Prince, based on a photograph of Goldsmith that the magazine Vanity fair had licensed it to Warhol. Warhol used the photo for an entire series—which Goldsmith only learned about when the magazine reused the screenprint image for a commemorative issue after Prince's death.
On the surface, this seemed to cast High Court judges as modern art critics, able to judge all appropriation art as transgressive. But the case was not whether Warhol's screen print inherently infringed Goldsmith's copyright, but whether he was infringing it for licensed use by a magazine in a way that could compete with the original photograph. Ultimately, there was a limit to transformative use. “The same copy,” the court ruled, “may be fair when used for one purpose but not for another.”
So it may be fair use for Google to copy entire books in order to create a searchable database of those books with excerpts from them, as it did for Google Books — but not necessarily for Suno or Udio to copy terabytes of recordings to stimulate the creation of new works to compete with them, especially if it results in similar works. In the first case, it's hard to find any real economic damage – there will never be a large market for licensing book databases – but there is already a nascent market for licensing music to train AI programs. And, unlike Google Books, the AI programs are designed to create music to rival the recordings used to train them. Obviously, licensing music to train an AI program is what we might call secondary use – but so is turning a book into a movie, and no one doubts that a license is needed for that.
All of this might sound like I think the major labels will win their cases, but that's a tough call — the truth is, I just don't think they will. I lose. And here there is a lot of space between victory and defeat. If one of these cases makes it to the Supreme Court — and if one of them doesn't, another case about AI training surely will in the next few years — the decision may be more limited than what the US seeks. either side, as the court tends to tread lightly around technological issues.
It is also possible that the decision will depend on whether the results resulting from all this training are sufficiently similar to copyrighted works to be labeled or reasonably labeled as infringing. Both label lawsuits are full of such examples, presumably because that could make all the difference. These cases concern the legality of AI inputs, but the determination of fair use in this matter could easily involve whether those inputs lead to infringing production.
In the end, Ginsburg suggests, “system designers may need to disable features that would allow users to create recognizable copies.” Besides – let's face it – isn't that really part of the fun? Sure, creating music with artificial intelligence may eventually mature into something of an art form—it already has enormous practical value for songwriters—but for ordinary consumers it's still hard to beat Frank Sinatra sings “Get Low” by Lil Jon. Of course, this could put significant strain on AI companies—with serious consequences for crossing a line that won't always be obvious. It might be easier for them to just license the content they need. The next questions, which will be the subject of future columns, are exactly what they should license and how they might do that, as it won't be easy to get all the rights they need — or in some cases even agree to who controls them.
from our partners at https://www.billboard.com/pro/ai-lawsuits-labels-andy-warhol-prince-1981-photograph-explained/
