Sony and Warner Sue Anthropic Over AI Training, Seeking Billions in Copyright Damages

Sony Music Publishing and Warner Chappell Music have sued Anthropic, accusing the artificial intelligence company of illegally using copyrighted music to train its Claude AI models.

The lawsuit, filed in the US District Court for the Northern District of California on August 28, names Anthropic as well as CEO Dario Amodei and co-founder Benjamin Mann as defendants. The publishers describe the alleged conduct as one of the largest and most blatant cases of ongoing intellectual property theft.

Anthropic has disputed the allegations and says it intends to defend itself.

The case could become one of the most important copyright battles yet between the music industry and an AI company because it goes beyond a dispute over individual songs. At its centre is a much bigger question:

Can an AI company use copyrighted creative works to develop its technology without obtaining permission from the people who own those works?

What are Sony and Warner accusing Anthropic of doing?

The music publishers allege that Anthropic obtained large amounts of copyrighted material through unauthorised downloading and scraping and used that material to train Claude.

According to the lawsuit, the material included copyrighted song lyrics.

The publishers argue that Anthropic's AI models can reproduce lyrics from copyrighted songs, which they say demonstrates that protected works were incorporated into the company's systems without the necessary permission.

This is important because AI companies need enormous amounts of data to train increasingly capable models.

For a system such as Claude, that training can involve processing huge collections of text and other material. The legal dispute is over where the line should be drawn when some of that material is protected by copyright.

In very simple terms: what is the dispute?

Imagine a songwriter spends years creating thousands of songs.

An AI company then obtains copies of those works and uses them to train an AI system that can generate text resembling or reproducing parts of those songs.

The question becomes:

Does the AI company need permission to use those works for training?

And if it does, who should be paid?

That question has become one of the biggest unresolved legal issues surrounding generative AI.

The music industry argues that its catalogues are valuable intellectual property and should not simply become free training material for AI companies.

AI companies, meanwhile, face the challenge of developing models that require enormous datasets while navigating copyright laws that were not written specifically for modern AI systems.

How much money is at stake?

The publishers are seeking potentially enormous damages.

The lawsuit seeks statutory damages of up to $150,000 for each infringed work, as well as damages connected to alleged removal of copyright-management information.

Because the complaint involves tens of thousands of copyrighted works, the potential financial exposure could reach billions of dollars.

That does not mean Anthropic will necessarily have to pay billions.

The final amount, if the case ultimately results in damages, would depend on what the courts determine Anthropic actually infringed and what remedies apply.

This is not Anthropic's first copyright fight

The new lawsuit adds to a growing series of legal battles surrounding Anthropic's use of copyrighted material.

Universal Music Publishing Group, Concord Music Group and ABKCO previously sued Anthropic over the alleged use of copyrighted song lyrics. That case involved roughly 500 songs initially and was later expanded into litigation involving more than 20,000 songs.

Sony Music Publishing and Warner Chappell joining the fight means the publishing arms of all three major music companies are now pursuing litigation against Anthropic.

Other publishers, including Round Hill Music, have also taken legal action against Anthropic.

So this latest lawsuit is not an isolated dispute.

It is part of a much broader confrontation between the music business and companies developing generative AI.

Why are music publishers taking such a hard line?

For music publishers, the issue is not simply whether AI can generate lyrics.

Their concern is control over the underlying creative works.

Songwriters and publishers earn money by licensing copyrighted compositions for uses such as recordings, performances and other commercial activities.

AI introduces another potential use: training the technology itself.

If AI companies can use large catalogues of copyrighted works without licensing them, publishers fear that an important new market could develop without compensating the people who created the underlying material.

That is why licensing has become such a central issue in the AI and music debate.

There is another problem: AI can reproduce copyrighted material

The publishers' case also focuses on what Claude can produce.

The lawsuit alleges that Anthropic's systems can generate lyrics that are identical or substantially similar to copyrighted lyrics.

That distinction matters.

Using copyrighted material during training is one legal question.

Having an AI system reproduce protected material in response to a user's request creates another.

The courts will ultimately have to consider how copyright law applies to both sides of that problem.

Why this case matters beyond Anthropic

The outcome could have consequences well beyond Claude.

AI companies across the industry rely on enormous datasets to develop their models. At the same time, publishers, authors, artists, photographers, software developers and other creators increasingly argue that their work should not be used to build commercial AI systems without permission or compensation.

That means a major court decision involving Anthropic could influence how future disputes are handled across the wider AI industry.

For AI companies, the stakes include potentially significant licensing costs and changes to how training data is acquired.

For copyright holders, the stakes include whether their existing rights remain meaningful in an industry increasingly built around large-scale data collection.

The music industry's strategy is changing

The lawsuit comes as major music companies pursue several different approaches to generative AI.

Some are fighting AI companies in court.

Others are investing in AI businesses and exploring licensing arrangements.

Sony, Universal and Warner, for example, recently participated in a $76 million investment round for Stability AI, showing that the relationship between the music industry and AI is not simply one of opposition.

The message appears to be more complicated: music companies want to participate in the AI economy, but they also want control over how their intellectual property is used.

What happens next?

The lawsuit will now move through the US legal system, where Anthropic will have an opportunity to respond to the allegations.

The central issues will include whether Anthropic infringed the publishers' copyrights, how its alleged use of copyrighted material should be treated under US copyright law and what damages, if any, should apply.

Those questions could take years to resolve.

But the direction of the dispute is already clear.

The music industry is no longer treating AI training as a purely technical issue. It is increasingly treating access to copyrighted creative works as a licensing and ownership question.

The bigger picture

The fight between Sony, Warner and Anthropic is ultimately about more than song lyrics.

It is about who gets to benefit when artificial intelligence learns from human creativity.

AI companies need data to build increasingly powerful systems.3

Creators and copyright owners want control over the works that make that training possible.

The legal battle now unfolding will help determine whether those two interests can coexist under existing copyright rules, or whether the AI era will force the rules themselves to change.