The recording industry has significantly escalated its tone against generative AI music. Sony Music Entertainment has filed a massive lawsuit against Udio, an AI-powered music generation startup, listing over 30,000 songs that it alleges were used to train the company's model without proper licensing.
The list of allegedly infringed works is extensive and eclectic, including artists ranging from Beyoncé and Britney Spears to Johnny Cash, Harry Styles, and dozens of other top-tier names across multiple genres and decades. What sets this lawsuit apart from earlier actions against AI music companies — such as the RIAA's 2024 lawsuits against both Suno and Udio — is its scale and evidentiary detail. Sony did not file generic accusations: it listed works one by one, building a concrete case that will be difficult to contest without Udio revealing details of its training dataset.
This move by Sony represents a qualitative escalation in the battle between the recording industry and generative AI. Unlike earlier lawsuits that questioned the general principle of training on protected data, Sony is operating on the terrain of material evidence. The message is clear: here are 30,000 protected songs that we believe are in your dataset, and here are generated recordings produced by your model that resemble them.
For Udio, a startup that built its product around the ability to generate complete songs from text descriptions, this lawsuit represents an existential threat. Unlike companies such as Anthropic with billions in cash reserves to absorb settlements, Udio operates on limited venture capital and may simply not survive a lengthy legal battle — or an equally expensive settlement.
The message the recording industry is sending is unequivocal: it will not passively accept the unlicensed appropriation of its catalog for training AI models. Any company that wants to build music generators will need to formally negotiate licenses or face detailed, expensive lawsuits. The era of silent AI training on copyrighted music data is, by all indications, coming to an end.
This case also signals a broader shift: copyright holders are becoming more sophisticated in how they pursue AI-related claims. Rather than relying on broad principles, they are investing in the forensic work needed to identify specific infringing uses. For AI companies in the music space, the window for operating without explicit licenses is closing rapidly.
Source: The Verge
