Like many AI companies, music startups Udio and Suno appear to have relied on unauthorized cuts of copyrighted works to train their models. This is by the admission of their own and investors as well new lawsuits filed against them by music companies. If these lawsuits go before a jury, the trial could be both damaging exposure and an extremely useful precedent for similarly sticky-fingered AI companies facing specific legal risk.
The lawsuits, filed by the Recording Industry Association of America (RIAA), put us all in the uncomfortable position of rooting for the RIAA, which for decades has been the daddy of digital media. I myself have received unpleasant from them! The case is simply that clear.
The gist of the two lawsuits, which are remarkably similar in content, is that Suno and Udio (strictly speaking, Uncharted Labs doing business as Udio) indiscriminately ransacked more or less the entire history of recorded music to form datasets, which then used to train an AI that produces music.
And here let’s quickly note that these AIs don’t “generate” so much as match the user’s prompt with patterns from their training data and then try to complete that pattern. In a way, all these models do is perform covers or mashups of the songs they received.
That Suno and Udio took the copyrighted data in question seems, for all intents and purposes (including legal ones), highly likely. The companies’ leadership and investors were unconsciously relaxed about the space’s copyright challenges.
They have admitted that the only way to build a good music production model is to consume a large amount of high quality music. It is quite simply a necessary step in building machine learning models of this type.
Then they said they did it without the record labels permission. Investor Antonio Rodriguez of Matrix Partners told Rolling Stone just a few months ago:
Honestly, if we had label deals when this company started, I probably wouldn’t have invested in it. I think they should have made this product without limitations.
The companies told RIAA lawyers that they believe the media it has swallowed falls under the fair use doctrine — which essentially only comes into play in the unauthorized use of a work. Now, fair use is admittedly a complex and murky concept in concept and execution, but the companies’ usage seems to fall outside the intended safe harbor of, say, a seventh grader using a Pearl Jam song in the background of a video them to global warming.
To be honest, it seems that the goose of these companies is cooked. Perhaps they hoped they could take a page from the OpenAI playbook, using evasive language and misdirection to stop their less serious critics, such as writers and journalists. (If once the AI companies’ skulduggery is exposed and it’s the only option for distribution, it no longer matters.)
But it’s harder to pull off when you have a smoking gun in your hand. And unfortunately for Udio and Suno, the RIAA says in its lawsuit that it has a few thousand smoking guns, and that the songs it owns clearly surpass music models. His claim: that whether Jackson 5 or Maroon 5, the “generated” songs are slightly scrambled versions of the originals – something that would be impossible if the original wasn’t included in the training data.
The nature of LLMs—namely, their tendency to hallucinate and lose the plot the more they write—precludes the regression of, say, entire books. This has likely led to a lawsuit by the authors against OpenAI, as the latter can reasonably claim that the passages cited by its model were taken from reviews, first pages available online, and so on. (The last move in the goal post is that they He made use copyright works early on, but they’ve since stopped, which is funny because it’s like saying you only got juice from an orange once, but you’ve since stopped.)
What you slope do it’s reasonable to claim that your music generator only heard a few bars of “Great Balls of Fire” and somehow managed to spit out the rest word for word and chord for chord. Any judge or jury would laugh in your face, and with luck a courtroom artist will have the opportunity to depict this.
This is not only intuitively obvious but also legally consistent, as the re-creation of entire works (confused, but apparently based on the originals) opens a new avenue for relief. If the RIAA can convince a judge that Udio and Suno are doing real and substantial harm to the business of copyright holders and artists, it can ask the court to shut down the entire operation of the AI companies at the start of the trial by injunction.
Are you opening paragraphs of your book coming out of an LLM? This is an intellectual issue that needs to be discussed at length. Dollar Store “Call Me Maybe” Created on Demand? Close it. I’m not saying it’s right, but it’s possible.
The predictable response from the companies was that the system is not are intended to reproduce copyrighted works: a desperate, bare-bones attempt to offload responsibility to users under the section 230 safe harbor. That is, in the same way that Instagram is not liable if you use a copyrighted song to support your Reel. Here, the argument seems unlikely to gain traction, in part because of the aforementioned admissions that the company itself ignored copyright in the first place.
What will be the consequence of these lawsuits? As with all things artificial intelligence, it is completely impossible to say in advance, as there is little prior or applicable, established dogma.
My prediction is that companies will be forced to expose their data and training methods, as these things have clear evidentiary interest. And if this evidence shows that they are indeed misusing copyrighted material, we will see an attempt to settle or avoid trial and/or a speedy judgment against Udio and Suno. It’s likely that at least one of the two will try to keep up, using legal (or at least legal-adjacent) music sources, but the resulting model will (by their own standards for training data) almost certainly take a huge step lower quality, and users will flee.
Investors? Ideally, they’ll lose their shirts, having placed their bets on something that was likely illegal and certainly unethical, and not just in the eyes of armed writers’ clubs, but according to the legal minds of the infamously and mercilessly litigious RIAA.
The implications can be far-reaching: If investors in a new new media production company suddenly see a hundred million dollars evaporate due to the fundamental nature of the production media, a different level of diligence will suddenly seem appropriate.
Companies may learn from trial or settlement documents what can be said — or perhaps more importantly, what shouldn’t be said — to avoid liability and keep copyright owners guessing.
While this particular suit seems almost preordained, it won’t be a playbook for prosecuting or picking off settlements from other AI companies, but an object lesson in infamy.
It’s good to have one of these every now and then, even if the teacher is the RIAA.
