US music industry groups file a brief saying AI training is not fair use in the OpenAI authors and newspapers cases

Five US music trade groups have asked a federal court to reject the argument that training AI on copyrighted writing is fair use. The RIAA, NMPA, A2IM, SONA and the Artist Rights Alliance filed an amicus brief in the lawsuits that book authors and newspaper owners have brought against OpenAI. Complete Music Update names those five as signers; other coverage lists more, including ASCAP and BMI, but that wider list is not confirmed. An amicus brief comes from outside parties who want the court to hear their view on a question the main litigants are already arguing; here that question is whether copying works to train a model is a lawful fair use. The groups state their position flatly: in their view AI training should never count as fair use, rather than being judged case by case.

The brief leans on the Supreme Court's 2023 ruling in the Andy Warhol Foundation case, under which a use counts as transformative only if it serves a purpose different from the original's. The groups say a court has to ask what the specific purpose was in copying each book or article, and they argue that AI models are trained to turn out stand-ins that read and sound like the works they learned from. In effect, they want judges to examine each use of each work rather than treat AI training as one abstract activity.

“affirmatively undermines the constitutional incentives for humans to author creative works”

the music organisations' amicus brief, on AI training, as printed by Complete Music Update, in the music groups' amicus brief · completemusicupdate.com ↗

A second argument is market dilution. AI can generate content without limit, the brief says, which displaces the market for the works it was trained on and drags down their value. Their reasoning is that a model able to produce unlimited material competes with the books and articles it absorbed, so the harm falls on the market as a whole and not only on one author. The groups present this as the factor that most favours rightsholders, and point out that a judge in the earlier authors case against Meta said such arguments could defeat an AI fair use defence. A third argument holds that AI companies pay little attention to the Supreme Court's instruction to weigh copyright's wider purpose, which is to encourage human creativity.

The groups also push back on the US Department of Justice under Donald Trump, which told the judge that a ruling against OpenAI would be at odds with basic copyright principles and hamper progress in science and the arts. In the music groups' telling, AI developers copy works to make digital imitations, which wears away the incentive copyright law gives people to create new work. The brief rejects the department's position without naming it. Other briefs, including the government's, back the tech companies' view that scraping for AI is fair use.

Music has its own stake. The brief notes that the labels' and publishers' suits against Suno, Udio and Anthropic rest on the same fair use question, and that every US AI copyright case turns on it, which is why the groups weighed in on a case that is not about music. The authors' and newspapers' cases against OpenAI are being heard together, so briefs filed in one are aimed at the same fair use question as the other.

A separate brief drew a different set of music names. The Music Artists Coalition did not sign the music groups' filing; it joined a 31-page brief led by Cate Blanchett's Human Consent Foundation in the New York Times case against OpenAI and Microsoft, filed on a Friday in the Southern District of New York under case number 1:23-cv-11195. Signers include Bette Midler, Helen Mirren, Julia Roberts, Robert De Niro, Rosario Dawson, Scarlett Johansson, Vince Gilligan and the Stan Lee Foundation. Digital Music News reported that filing on 1 November. The foundation began life as RSL Media in May and opened a registry in June where people log what AI permissions they give over their likeness.

OpenAI and Microsoft, as defendants, have argued that licensing creative work for AI training costs too much to be practical. The Human Consent Foundation's coalition says that claim is wrong, because creative industries already run rights-clearing systems, so the machinery for paying creators exists and cost is no reason to excuse copying. Those systems, which in music include labels, distributors, publishers, collecting societies, Merlin and IMPEL.