Sony Music Entertainment has filed another lawsuit against the AI music generator Udio, alleging the startup illegally copied more than 30,000 songs to train its models. Sony says the alleged material comes from artists including Beyoncé, Harry Styles, and Elvis Presley, which is not just a brand problem for Udio. It is a legal and operating problem: if the training set is treated as unlawfully sourced, the entire model development pipeline can become exhibit A.

This “again” matters. Sony first brought lawsuits against Udio and its competitor Suno in June 2024, in partnership with major labels Universal Music Group and Warner Music Group. That means the current filing is not happening in a vacuum. It is part of a continuing enforcement campaign aimed at changing the incentives for how AI music companies obtain, process, and claim rights to the audio data they use.

For executives watching from outside the courtroom, the underlying tension is straightforward: training AI typically depends on large datasets, but music rights are locked behind copyright and licensing structures. The dispute is not merely about whether AI can create music. It is about whether the system was trained with permission or through infringement. When a label like Sony frames the allegations around “copying” tens of thousands of songs, the stake shifts from abstract policy to concrete damages and injunctive relief, which can directly affect product timelines, licensing costs, and the legal viability of certain training approaches.

There is also a strategic communications layer. Mentioning major artists such as Beyoncé, Harry Styles, and Elvis Presley is a deliberate move to make the claims legible to both courts and consumers. In practice, these cases tend to become high-scrutiny battles over what “use for training” means legally, whether a company can rely on any defenses, and how courts should weigh the rights of copyright holders against the public interest in innovation. That is why the timeline is important. Sony acted in June 2024, escalated with additional litigation, and is now pressing again, signaling it expects progress through continued legal pressure.

Industry dynamics make the repetition even more consequential. The AI music space is crowded with new entrants and fast-moving product cycles, so the legal map can change quickly. If courts or settlements impose constraints on training data, companies may need to pivot toward licensed datasets, partnerships with rightsholders, or new technical methods that reduce reliance on potentially unapproved sources. Even if an individual lawsuit does not end the fight, litigation creates compliance overhead: legal reviews of training corpora, documentation of provenance, and tighter control of data pipelines. Those are board-level issues because they affect valuation, runway, and future fundraising.

For decision-makers, the second-order risk extends beyond the two companies named in the original June 2024 suits. Sony’s involvement alongside Universal Music Group and Warner Music Group suggests a coordinated label view of AI music threats and remedies. When multiple major labels align, executives should assume they are also establishing a pattern they intend to enforce across the ecosystem. That can influence how investors price legal risk for AI generative media, and how platforms set their own policies for model training and output handling.

The most practical takeaway is that “training” is becoming the battleground. In AI more broadly, many disputes have focused on outputs, similarity, or use cases. Here, Sony is alleging the root issue is the training process itself, centered on more than 30,000 songs. If that allegation sticks in any meaningful way, it can force a rethink of how music AI models are built and what evidence companies must maintain to show lawful sourcing. For the wider market, that could accelerate licensing conversations, reshape product roadmaps, and raise the price of noncompliance.

In short: Sony is not just testing Udio. It is testing the boundaries of how the industry digitizes music for machine learning, and how quickly rightsholders can turn those boundaries into enforceable outcomes. If you are building, funding, or partnering in AI music, you should treat this as a live signal: the legal fight is ongoing, the dataset question is central, and the consequences for operations and strategy can arrive faster than most teams expect.