The legal battle over how artificial-intelligence companies train their systems is expanding into the music industry, with independent publisher Round Hill Music filing separate copyright lawsuits against Anthropic and Suno and seeking damages that could eventually exceed $1 billion.

Round Hill alleges that the two AI companies used copyrighted song lyrics without authorization to train their systems. The publisher says the cases currently cover more than 500 compositions and that the claims could be expanded to include more than 10,000 works from its catalog.

The lawsuits, filed in federal court in Northern California, add two more major defendants to a rapidly expanding dispute over whether copyrighted material can legally be copied and processed at scale to build commercial AI products.

The dispute is part of a much broader confrontation between copyright owners and technology companies.

Generative AI systems need enormous quantities of information for training. Developers argue that ingesting copyrighted material can be comparable to learning from publicly available works and that the training process can fall within doctrines such as fair use.

Copyright owners take a very different position.

They argue that AI companies are taking valuable creative works, using them to build commercial products and potentially generating enormous revenues without compensating the people who created or own the underlying material.

Music publishers have particular concerns because lyrics represent highly protected creative works and because generative AI systems can produce outputs that resemble existing musical styles and compositions.

Round Hill is now putting those questions directly before federal courts.

The case against Anthropic centers on the company's Claude AI system.

Round Hill alleges that copyrighted lyrics were copied and used in training the model without permission. The publisher's catalog includes works associated with artists such as James Brown, The Kinks and the Goo Goo Dolls.

The significance of the claim goes beyond the individual songs.

If courts determine that copying protected lyrics into an AI-training process constitutes infringement, AI developers could face much larger licensing obligations.

That could change the economics of developing large language models.

Technology companies might be required to negotiate licenses for extensive collections of books, music, articles and other copyrighted material before they can legally use those works as training data.

Round Hill's lawsuit against Suno focuses on its AI-powered music-generation platform.

Suno allows users to generate new music using artificial intelligence, making the copyright issue especially direct.

The company is not simply using AI internally for search or analysis. It sells a service that creates commercially valuable musical output.

That makes the question of how the model learned to generate music particularly important.

Music labels and publishers have already brought legal action against Suno and other AI-music companies, arguing that unauthorized use of recordings, compositions or lyrics could undermine the economic value of human-created music.

Round Hill's potential damages claim could exceed $1 billion if the litigation is expanded to more than 10,000 compositions.

That figure is important, but it should not be interpreted as money the companies have been ordered to pay.

The lawsuits are at an early stage, and the plaintiffs will still have to establish the relevant copyright violations and overcome defenses that Anthropic and Suno may raise.

The eventual damages, if any, could be far lower.

Even so, the size of the claim demonstrates the financial stakes.

A major ruling against AI companies could encourage thousands of additional copyright owners to bring similar cases.

Round Hill's lawsuits join an expanding list of copyright disputes involving AI developers.

Anthropic has already faced litigation from major music companies and other copyright holders, while Suno is also defending separate cases brought by major music-industry companies.

The legal pressure is not limited to music.

Authors, publishers, visual artists and software developers have all questioned whether AI companies can lawfully ingest their work without permission.

The cases could eventually produce a more standardized framework for AI training.

At the heart of the dispute is an unusually difficult legal question.

Humans learn from copyrighted works without obtaining a license every time they read a book or listen to a song. AI developers argue that models similarly learn statistical patterns from enormous datasets rather than storing conventional copies for later resale.

Copyright owners counter that AI training requires making copies during the process and that those copies are part of the commercial activity that creates the AI product.

Courts will therefore have to examine not simply what an AI model ultimately produces, but how the system was created.

That distinction could prove critical.

Round Hill has made clear that it does not want AI companies to receive unrestricted access to copyrighted music for free.

The broader music industry increasingly sees licensing as a potential way forward.

A licensing model could allow AI companies to continue developing powerful systems while giving publishers and artists a financial return when their works are used to build those systems.

The challenge is determining how such a system could operate at the enormous scale required by modern AI.

A model may be trained on millions or even billions of pieces of information. Negotiating individual permissions for every work could be difficult, expensive and slow.

That is why the Round Hill lawsuits matter beyond the music business.

If licensing becomes mandatory, the cost of training AI systems could rise substantially.

Developers could have to negotiate large-scale data agreements with publishers, record companies, artists and other rights holders.

That would create a new market around AI training data.

On the other hand, if courts broadly accept fair-use arguments, AI developers could retain much greater freedom to use existing material, potentially lowering the cost of building new models.

The timing is particularly important because AI companies are moving rapidly into commercial applications.

Anthropic is building increasingly sophisticated software and enterprise systems, while Suno is attempting to turn generative music into a mainstream consumer product.

Both businesses depend on the underlying training process.

Round Hill's legal strategy therefore attacks an issue at the foundation rather than at the edges of AI development.

The cases will take time to work through the courts, and the ultimate outcomes remain uncertain.

But the message from music-rights holders is becoming increasingly clear: AI companies may not be able to build billion-dollar businesses from copyrighted creative works without eventually answering the question of who owns, licenses and gets paid for that material.

Round Hill's lawsuits could become another important test of that principle — and potentially one of the most expensive copyright battles yet between the traditional creative economy and the rapidly expanding AI industry.