RIAA’s past statements undermine current AI lawsuit claims, says Mike Masnick

Mike Masnick of Tech Dirt says that the RIAA’s previous statements contradict their stance on AI, potentially unraveling their latest legal battles. 

Op-ed Mike Masnick of Tech Dirt

There have been a bunch of lawsuits over the last couple of years from traditional content industries suing AI providers, claiming copyright infringement. We’re still a long way from figuring out how all of these lawsuits will shake out. We’ve made it clear that we’re skeptical of these lawsuits, largely because you would have to basically ignore a bunch of important and useful copyright precedents to reach the conclusion that training on copyright-covered works infringes on copyright.

Of course, this is copyright, where logic and precedent are often ignored based on who a judge hates more. So, we shall see. But, we’ve definitely seen a lot of people cheering on these lawsuits, mainly in the false belief that it’s about “artists” vs. “big evil tech companies” and therefore the “artists” should win.

Reality is always a lot more complex and nuanced. If these lawsuits succeed, it will not help artists get paid. Instead, it will again increase the reliance on middlemen who have a long history of screwing over the artists. Just the fact that the RIAA is currently run by a guy who famously got his job at the RIAA just months after sneaking language into a bill to fuck over musicians should tell you all you need to know about the RIAA’s actual interests.

Also, if the cases decide that training is a licensable scenario, it will kill smaller and open source AIs and make it so only the largest of the largest tech companies can create LLMs. So instead of being a victory over “big tech,” it will hand the market to big tech.

And that’s not even getting into the damage it would do to the ability to read the open internet (which itself could be judged a licensable event) or the ability of researchers to scan and collect data about the open internet.

Just be careful what you wish for.

Earlier this week, the RIAA gleefully announced that it was suing two of the bigger music generator AI services. It filed one lawsuit against Suno in Massachusetts and another against Udio in New York.

Both lawsuits are effectively the same. And, they’re both ridiculously weak. They are both based on the premise that training on copyright-covered works requires permission. But, again, we’ve been there and done that. Training is a form of scanning or reading, and that’s either not a copyright-triggering event at all, or it’s fair use.

The lawsuits do not name what copyright covered content was actually copied beyond some handwaving about “all of it.” This is not sufficient for a copyright claim. The lawsuits argue that because it can tell these apps to make songs like musicians on RIAA member labels, that proves it’s infringing. From the Suno complaint:

Plaintiffs could have proceeded with this action based solely on eliciting that reasonable inference of copying. Nevertheless, Plaintiffs’ claims are based on much more. In particular, Plaintiffs tested Suno’s product and generated outputs using a series of prompts that pinpoint a particular sound recording by referencing specific subject matter, genre, artist, instruments, vocal style, and the like. Suno’s service repeatedly generated outputs that closely matched the targeted copyrighted sound recording, which means that Suno copied those copyrighted sound recordings to include in its training data. In addition, the public has observed (and Plaintiffs have confirmed) that even less targeted prompts can cause Suno’s product to generate outputs that resemble specific recording artists and specific copyrighted recordings. Such outputs are clear evidence that Suno trained its model on Plaintiffs’ copyrighted sound recordings.

Which… doesn’t matter? Again, training is clearly fair use, and “specific subject matter, genre, artist, instruments, vocal styles, and the like are not copyright-covered expression. All of those things are not elements subject to copyright.

If you want proof of that, just look at what the RIAA itself has said in cases a few years ago, following the Blurred Lines decision that initially suggested that music “styles” should be covered by copyright. The RIAA realized, quite quickly, that this might make a huge portion of the labels’ catalogues infringing and freaked out. In one case, the RIAA filed an amicus brief noting that such overprotection would be incredibly damaging.

… new songs incorporating new artistic expression influenced by unprotected, pre-existing thematic ideas must also be allowed.

That’s the RIAA’s own argument just six years ago. And now they’re arguing that such unprotected thematic ideas are protected. But only when tech companies are making use of them apparently.

Again, in that brief, the RIAA cogently argues against what the RIAA is now arguing in these complaints:

Most compositions share some elements with past compositions—sequences of three notes, motifs, standard rhythmic passages, arpeggios, chromatic scales, and the like. Likewise, all compositions share some elements of “selection and arrangement” defined in a broad sense. The universe of notes and scales is sharply limited. Nearly every time a composer chooses to include a sequence of a few notes, an arpeggio, or a chromatic scale in a composition, some other composer will have most likely “selected” the same elements at some level of generality.

To keep every work from infringing — and to keep authors from being able to claim ownership of otherwise unprotected elements — this Court has stressed that selection and arrangement is infringed only when there is virtual identity between two works, not loose resemblance. The same principle should be recognized for music.

Um. So, considering that the complaints are not showing “virtual identity between two works” then the RIAA itself has made the case for why these models are not infringing.

In that same brief, the RIAA itself admits that there can be only “thin” copyright coverage on general themes at most, to avoid making music inspired by others to be infringing:

To prevent nearly every new composition being at risk for liability, copyright claims based on “original contributions to ideas already in the public domain,” Satava v. Lowry, 323 F.3d 805 (9th Cir. 2003), are seen as involving a “thin copyright that protects against only virtually identical copying.” Id. at 812; see also Ets-Hokin v. Skyy Spirits, Inc., 323 F.3d 763, 766 (9th Cir. 2003) (“When we apply the limiting doctrines, subtracting the unoriginal elements, Ets-Hokin is left with . . . a ‘thin’ copyright, which protects against only virtually identical copying.”); Rentmeester v. Nike, Inc., 883 F.3d 1111, 1128-29 (9th Cir. 2018). This Court has long recognized this principle in claims involving visual art that allegedly creatively combines public domain elements, as with the sculptures in Satava or the photographs in Ets-Hokin and Rentmeester. The same should apply to music.

Perhaps Suno and Udio should take a page from the RIAA’s own legal arguments in responding to these complaints against them.

I am sure RIAA folks (and anti-AI folks) will rush in to explain why “this is different,” but it’s not. It’s literally the same argument. Does copyright actually protect genre, themes, and the like? Of course not. It would be a ridiculous and dangerous outcome should that come to pass.

Now, I know the RIAA will claim that it’s not suing over the output of these tools, but rather just pointing to those things as proof of infringement on the training side. But, again, training by scanning copyright-covered material for a totally transformative use (which includes learning from or being inspired by) is quintessential fair use.

The training is fair use. The fact that it can output songs with a similar theme matters not one bit to the copyright question, as the RIAA itself admits.

Of course, this case will go on for years and years. You can never predict how courts will rule on copyright issues, but these two cases seem particularly weak and silly. This is especially true given how it shows the RIAA going back on its own previous claims from just a few years ago.

And, just to close out this piece, I’ll note that RIAA CEO Mitch Glazer, again, the very guy who snuck words into a totally unrelated bill to literally take copyrights away from artists and hand them to music labels, is quoted in the press release about this lawsuit talking about how it’s not fair to “exploit” an “artist’s life’s work” for profit, even though that’s exactly what all of his member labels have done for nearly all of their existence.