The fight over AI music generators has moved from industry debate to federal court. The Recording Industry Association of America (RIAA), Universal Music Group, Warner Music Group, and Sony Music have filed lawsuits against Suno and Udio, accusing the AI music companies of massive copyright infringement.
At the center is a question that reaches beyond music: can a generative AI model be trained on copyrighted works without permission if the company argues that the model learns from them rather than copies them?
What the labels allege
The labels say Suno and Udio illegally copied copyrighted sound recordings to train their AI models. According to Billboard, the major labels accuse the startups of infringing on copyrighted sound recordings "at an almost unimaginable scale."
The RIAA said lawsuits have been filed against Suno, Inc., developer of Suno AI, in federal court in Massachusetts, and against Uncharted Labs, Inc., developer of Udio AI, in federal court in New York.
The complaint is not only about outputs that may sound like existing songs. The deeper allegation is that the services used protected recordings without permission to build the systems that generate music in the first place.
The plaintiffs argue that the goal is to create music that would "saturate the market with machine-generated content that will directly compete with, cheapen and ultimately drown out the genuine sound recordings on which [the services were] built."
The labels are seeking an injunction to stop the companies from using copyrighted songs for training purposes, along with damages for existing copyright infringements.
Why the case matters for AI training
This dispute resembles other lawsuits against makers of large image, voice, and code models. In those cases, plaintiffs argue that copyrighted material was used without consent to train systems that can then compete with the original works or their creators.
Model makers often answer with a fair use argument. They claim that AI systems transform training data by learning patterns from it, rather than simply copying it. The comparison often made is that a person can study existing works and then create something new.
The music labels reject that framing. Their complaint says fair use applies in limited circumstances that promote human expression, while Suno and Udio produce machine-generated music imitations.
"The doctrine of fair use promotes human expression by permitting the unlicensed use of copyrighted works in certain, limited circumstances, but [the services] offe[r] imitative machine-generated music—not human creativity or expression."
That makes the lawsuit important for the broader generative AI industry. There is still no definitive ruling on fair use for this kind of AI training, and a case brought by major music companies against two smaller AI startups could help set a precedent.
Similarity claims add pressure
The fair use debate becomes more difficult when AI models produce content that appears to resemble protected works. Courts will need to determine whether those outputs are flaws in the models or part of their functionality, and whether their frequency could justify banning AI models or requiring royalties.
The RIAA has released examples of alleged copyright infringements due to similarity. It accuses Suno and Udio of specific plagiarism of songs by well-known artists, including Jason Derulo, Jerry Lee Lewis, and CashMoneyAP. Udio has already removed the tracks in question.
The source also notes that in April, AI developer Ed Newton-Rex accused Suno AI of strongly imitating well-known hits in style, melody, harmony, or instrumentation. Suno blocks prompts with artist names, but small spelling mistakes or precise descriptions of a certain style can bypass this protection.
That matters because the labels are not only arguing about what went into the models. They are also pointing to what can come out of them.
How Suno and Udio responded
Udio has issued a detailed response to the lawsuit. The company argues fair use, transformative use of data, and says Udio is a tool for creative people.
Udio says generative AI models learn from examples, comparing the process to students listening to music and studying scores. The company presents itself as part of a line of music innovations, including synthesizers and digital recording technology, that were initially met with skepticism but later advanced art and business.
Udio also says its technology is aimed at creating new musical ideas, not reproducing copyrighted works or artists' voices. The startup states that "we are completely uninterested in reproducing content in our training set."
However, the key criticism in the lawsuit is not reproduction alone. It is the alleged use of unlicensed data to gain music generation skills. The source notes that Udio's response does not directly address the content of its training sets or how it acquired that data, and that the word "data" is not mentioned in Udio's statement.
Suno AI has not yet commented on the lawsuit.
The licensing question ahead
The RIAA says the music industry generally welcomes AI and works with "responsible developers" to use it as a tool for artists. RIAA Chairman Mitch Glazier argues that services like Suno and Udio, operating without a license, would hinder progress.
The RIAA's position is direct: AI companies must obtain permission from rights holders and cannot claim fair use when AI-generated content does not represent human creativity. It also argues that unauthorized use of copyrighted music for AI training threatens the value of protected recordings.
The source points to Jen as a different model for AI audio generation. According to its creators, Jen is trained only on licensed content and pays royalties to artists.
For now, the Suno and Udio cases concentrate many of the unresolved questions around generative AI into one clear conflict: what AI companies can learn from, what they can generate, and whether the law treats those two steps separately or as parts of the same copyright problem.