A group of music industry organisations - including RIAA, NMPA, A2IM, SONA and ARA - have intervened in the big ongoing legal battle between a plethora of book authors and newspaper owners and Open AI.
In a so-called ‘amicus brief’ - filed with the court overseeing that legal battle - the music organisations have set out the music industry’s current arguments for why AI training should never be considered fair use under American copyright law. If AI training is fair use, that would mean AI companies could use existing works to train their models without getting permission from any creators or rightsholders.
Although this particular legal battle centres on Open AI’s copying of books and journalism in order to train its AI models, all of the many copyright lawsuits filed against AI companies in the US swing on the fair use argument - and that includes the music industry’s litigation targeting Suno, Udio and Anthropic.
Which is why what happens in this case involving authors and newspapers matters to the music industry, and why the music organisations have filed an amicus brief to set out their position.
This is also the legal battle where Donald Trump’s Department Of Justice previously told the judge that any decision by the court that said Open AI’s training was not fair use would be “inconsistent with basic copyright law principles” and “severely hamper” the “progress of science and arts”. In other words, “make sure you don’t screw up the USA’s leading position in the AI race with pesky copyright rules”.
Given the DoJ’s intervention on the side of Open AI, copyright industries like the music industry definitely feel the need to back the authors and newspaper publishers in this case.
Although they don’t criticise Trump’s DoJ directly, the music organisations very much reject the position it has taken in their new court filing. After all, they argue, the actual purpose of copyright law is “to encourage humans to produce creative works for the public’s benefit”.
Allowing AI companies to copy existing works without getting permission in order to “generate digital imitations” does the opposite of that, and in doing so “affirmatively undermines the constitutional incentives for humans to author creative works”.
The music industry’s amicus brief
The music industry’s amicus brief begins by noting the similarity between the Open AI legal battle and the various lawsuits filed by record labels, music publishers and musicians against music AI companies, and how, “like defendants in this case, the services targeted by the music community insist that training on their copyrighted works is fair use”.
When it comes to the AI companies’ fair use arguments, “they are wrong”, the court filing insists. And, because the members of these music industry organisations “own some of the most cherished copyrights in modern culture”, they feel it is important to lay out for the judge in this case the various “defects” they see in Open AI’s fair use defence.
The music industry’s resolute position that AI training is not fair use is now familiar to most people following the various lawsuits and lobbying campaigns, and we’ve already seen some of the key arguments against the fair use defence articulated in the Suno, Udio and Anthropic cases.
However, those arguments are being continually refined and reprioritised as the wider debate over the copyright obligations of AI companies evolves, so this new amicus brief provides a good summary of the music industry’s current thinking.
The music industry’s arguments
There are three main arguments presented in the new court filing.
First, the music organisations tackle the issue of “transformativeness”. One factor that US courts consider when weighing up a fair use defence in any copyright case is whether or not the use of a copyright protected work is ‘transformative’.
So if an existing work is copied to create a new work, is the new work ‘transformative’ and sufficiently different to the original? The more transformative a use is, the more likely it is to be fair use.
The AI companies have argued repeatedly that - because the content their models output is not exactly the same as any content used during training - that use is clearly transformative.
But, the music organisations say, the AI companies, including Open AI, are applying this test incorrectly. Citing the 2023 Supreme Court fair use ruling involving the Andy Warhol Foundation, they say that - when considering if a use is transformative - the court needs to consider whether the new work “serves a purpose akin to that of the original work”.
And, they add, the court must be specific when it considers the purpose of Open AI copying any one book or article, which is to say when it considers the reason Open AI needs to make the copy.
“An AI model does not ingest romance poetry or bluegrass to learn about language or music at large”, say the music industry’s filing, “but rather to generate substitutes that read and sound like them in practice, thus achieving the same, non-transformative purpose”.
The second argument presented in the music industry’s filing relates to the question of ‘market dilution’, which considers whether a new work created from by copying an existing work is likely to negatively affect the value of that original work.
This is another fair use factor US courts have to consider, and it’s the factor that probably most favours copyright owners in the AI debate - not least because in an earlier case involving book authors and Meta, the judge said market dilution arguments could defeat an AI fair use defence.
If the use of a copyright protected work can be shown to negatively impact on the market value of that work, that is less likely to be fair use. The music industry argues that when AI companies copy existing music to train generative AI models, which then output new music, that negatively impacts on the market value of the existing music.
And because of that market dilution, the AI company’s copying cannot be fair use. The authors and newspaper companies can make a similar argument.
Setting out this argument in their new court filing, the music organisations write that AI “has the unique capacity to output endless content that displaces the market for and diminishes the value of the copyrighted works on which it trained”. That kind of harm, they then argue, “falls squarely within the types of substitutional market impacts that defeat fair use”.
Third and finally, the music organisations present a big-picture argument against the AI sector’s fair use defence. Open AI - and all the other AI companies - they say, “pay little heed to the Supreme Court’s requirement that courts consider the broad purposes of copyright law when balancing the statutory fair use factors”.
Which means that, while there are specific fair use factors for the courts to review, judges also need to consider if the fair use principle is being relied upon in any one situation to help ensure the “broad purpose” of copyright is met. Which, the music organisations say, is “to encourage humans to produce creative works for the public’s benefit”.
When AI companies try to use the fair use principle to copy existing works to train AI models, they are not serving that purpose, the music organisations argue. They are not “stimulating human creativity”, but instead by “infringing copyrighted works to generate digital imitations” they are in fact undermining “the constitutional incentives for humans to author creative works”.
One more thing…
Although this main music industry amicus brief intervening in the Open AI case involves organisations representing labels, publishers, artists and songwriters, the US Music Artist Coalition has actually put its name to a different submission instigated by the Human Consent Foundation, a campaign group co-founded by Cate Blanchett.
Its submission includes another important point to counter another narrative of Open AI that Trump has shown sympathy towards - the idea that expecting AI companies to secure licences for every bit of copyright protected work they have copied is impractical and unrealistic.
“Defendants have argued that licensing creative works for AI training is cost-prohibitive”, the Foundation’s submission states. But “the facts demonstrate otherwise”, because, “the creative industries already have in place a proven infrastructure and workflows to clear rights for dispersed works at scale and cost-effectively”.
Which is to say there are licensing systems in place that AI companies could use to license the content they need for training.
For the music industry, that means the network of labels, distributors, publishers and collecting societies, and other licensing organisations like Merlin, IMPEL and ICE that bring together some of these entities. This network already allows streaming services to access vast quantities of music, and it’s there to do the same for AI companies that are willing to negotiate a deal.
Another co-founder of the Human Consent Foundation, Nikki Hexum, says she feels it was important to point this out, to show the court “how the creative industry actually works and acknowledge the solutions that have been built and are continuing to develop for this new digital era”.