Recently, a number of prominent record labels have initiated copyright infringement lawsuits in United States courts against the developers of the generative AI music applications known as Suno and Udio. The record labels assert that these AI companies unlawfully reproduced numerous sound recordings owned by them, leading to outputs that closely resemble the original tracks. In light of these allegations, the labels are pursuing damages amounting to $150,000 per infringed track, a figure that could accumulate to substantial sums given the thousands of tracks in question.
The litigation claims that Udio generated outputs strikingly similar to renowned songs such as ABBA’s “Dancing Queen” and Mariah Carey’s “All I Want for Christmas Is You.” Concurrently, Suno is accused of creating songs reminiscent of classics including James Brown’s “I Got You (I Feel Good)” and Chuck Berry’s “Johnny B. Goode.” This legal action underscores a growing trend of copyright disputes facing the rapidly emerging sector of generative AI technology, which has also seen visual artists and various newspapers take legal action against AI developers for similar concerns.
For readers unfamiliar with generative AI music applications, their functionality is quite straightforward. Users input a text prompt—such as “compose a female jazz song about beating the Monday morning blues”—and may also provide lyrics if they so wish. The application then utilizes its AI algorithms, having been trained on extensive datasets, to produce an MP3 of an original song that includes both vocal and instrumental elements, available for the user to download.
The crux of the allegations lies in the assertion that the data utilized to train these AI systems includes copyrighted sound recordings which were utilized without permission. The outcome of the legal proceedings will largely pivot on whether the actions taken by Suno and Udio in relation to these recordings can be classified as “fair use.”
In the United States, the fair use doctrine serves as a defense against copyright infringement claims. Conversely, Australia operates under a more restrictive “fair dealing” framework, applicable only to certain scenarios such as academic research and study. The determination of fair use will be assessed based on four specific factors: the purpose and character of the use, the nature of the original work, the volume and significance of the material used, and the potential market impact of the use on the original recordings.
The factor of greatest contention is likely to be the purpose and character of the use, which necessitates an evaluation of the generative AI’s transformative capacity—whether the output provides new meaning or value compared to the original compositions. The position of Suno and Udio rests on their argument that their technology is sufficiently transformative, as it synthesizes new and original output rather than merely reproducing existing songs.
The courts will also assess the amount and significance of the copied content, analyzing the role these recordings played in training the AI and in output generation. Notably, the consideration of substantiality may take a qualitative form; thus, the court will contemplate whether the reproduction of key elements of a song constitutes a violation, irrespective of the percentage of the work used.
Moreover, the implications for market value will also be scrutinized. The courts will determine whether the generative AI applications substitute original songs in a manner that significantly affects the market. This aspect is prone to interpretation from both sides of the legal argument.
An additional area of significant concern within the music industry pertains to the cloning of voices. While Suno has explicitly denied the capability for voice cloning within its app, citing restrictions on referencing specific artists, this matter will likely be addressed during the legal proceedings.
The trajectory of these lawsuits remains uncertain. It is possible that settlements may be reached prior to trial or that novel licensing arrangements could be established among the involved parties, reminiscent of OpenAI’s recent partnership with News Corp. However, with the advent of other innovative AI voice cloning products being championed by various startup companies, such as Hooky—a licensing platform for AI voice modeling that affords artists more control over the use of their voices—this landscape continues to evolve rapidly.
Should these lawsuits progress, American courts stand poised to provide clarity on whether the training methods and outputs produced by generative AI music applications fall within the purview of fair use. Such determinations could set vital precedents for the operational practices of various generative AI platforms department-wide.
Wellett Potter, Lecturer in Law, University of New England
This article is republished from The Conversation under a Creative Commons license.
For more information, please refer to the original source.
Leave a Reply