Musicians’ union pushes back against Universal and Warner’s AI licensing deals
The American Federation of Musicians is opposing Universal and Warner’s attempts to dismiss its lawsuit, arguing members’ recordings were used in AI systems without compensation.
The American Federation of Musicians (AFM) is continuing its legal fight against Universal Music Group and Warner Records, arguing that recordings performed by its members were licensed to AI companies without the musicians receiving compensation or being properly notified.
The union filed its opposition to the labels’ motions to dismiss in a New York federal court on September 4. At the centre of the dispute is whether AI licensing represents a “new use” of recordings under the AFM’s Sound Recording Labor Agreement (SRLA), potentially triggering payment and notification requirements for the musicians involved.
Musicians say their recordings were used to power AI
The AFM first filed its lawsuit in June, claiming Universal, Warner Records and Atlantic Recording Corp. breached their collective bargaining agreement by licensing recordings featuring AFM musicians to AI companies Suno and Udio.
In its latest filing, the union argues that these agreements allowed recordings made by its members to be “fed into AI systems for commercial exploitation” without the musicians being paid for that use.
The case is particularly significant because the labels themselves have previously argued that unlicensed AI training could harm human musicians. Universal, Warner and other major record companies sued Suno and Udio in 2024, claiming that AI-generated music could replace human-created recordings and dilute the royalty pool available to artists.
The AFM is now pointing to those arguments as evidence that the labels understand the commercial significance of allowing AI companies to access recorded music.
What does the ‘new use’ agreement mean?
The dispute comes down largely to Article 21(a) of the Sound Recording Labor Agreement. The provision covers what is described as a “new use” of a recording, essentially, using a recording for a purpose that isn’t already covered by the agreement.
The AFM argues that licensing recordings to AI companies for model training and music generation falls within this category. The union says the labels therefore had an obligation to notify the musicians and provide compensation.
Universal has argued that Article 21(a) does not establish its own payment rate and instead relies on rates contained in other agreements. Because there is no separate agreement specifically covering AI use, Universal says there is no applicable payment obligation.
Warner Records and Atlantic have made their own arguments for dismissing the case.
The AFM disagrees, saying the wording of the agreement is at least open to its interpretation and that this is enough for the case to continue rather than being dismissed at this stage. The union argues that it is asking the court to enforce the agreement as written, rather than create new terms.
Warner and Universal have already made AI deals
The dispute also highlights the rapidly changing relationship between major labels and AI music companies.
Warner settled with Udio in November 2025 and shortly afterwards became the first major label to settle with Suno. Universal reached a settlement and licensing agreement with Udio in October 2025, although it has continued its legal dispute with Suno.
Those deals have since become part of the wider shift towards licensed AI music. Suno, for example, has now launched its V6 generation of models using licensed music through partnerships including Warner Music Group and BMG. Warner has described its agreement with Suno as opening up new “creation-based” revenue opportunities for artists and songwriters.
But the AFM argues that the musicians whose performances helped create the underlying recordings should not be left out of those commercial arrangements.
According to the union, its members did not receive a share of the settlement proceeds from the labels’ AI agreements.
A bigger question for musicians and AI
The case could have important implications beyond the specific agreements between the AFM and the major labels.
AI companies are increasingly moving towards licensing music rather than relying solely on disputed training practices. At the same time, questions remain over exactly who should be compensated when recordings are licensed for AI development, how performers should be credited and whether existing contracts adequately cover technologies that did not exist when those agreements were negotiated.
That issue is becoming increasingly relevant for independent artists, session musicians and producers too. As AI companies develop systems capable of generating music, recreating styles and working with licensed recordings, the distinction between a recording being used as a traditional piece of music and being used as training or creative input for an AI system is becoming increasingly important.
The AFM’s lawsuit is effectively asking whether existing performer protections can already cover this new technology, or whether new agreements need to be negotiated specifically for AI.
The union has also pointed to the labels’ own history of negotiating new provisions when disputes emerged over areas such as video games and foreign licensing. It argues that this demonstrates that questions around the reach of existing agreements have previously been resolved through bargaining rather than simply excluding new uses altogether.
The case is still ongoing
The AFM’s opposition does not decide whether the musicians ultimately have a right to payment. Instead, it is arguing that the lawsuit should be allowed to proceed rather than being dismissed before the underlying issues can be examined.
The union and the labels are also negotiating the next version of their Sound Recording Labor Agreement, with AI reportedly a major issue in those discussions.
For musicians, the case is another sign that the debate around AI music is moving beyond the question of whether AI companies can legally train their models on copyrighted recordings. Increasingly, the industry is also having to work out what happens after those recordings are licensed, and whether the performers whose work helped create them are entitled to a share of the resulting value.
As AI becomes more deeply embedded in the music business, those questions could prove just as important as the copyright battles currently playing out between labels and AI companies.