Sony Music Publishing and Warner Chappell Music have launched a major copyright battle against artificial intelligence company Anthropic, accusing it of using copyrighted songs, lyrics and musical compositions without permission to train its Claude AI models.

The lawsuit, filed in a federal court in Northern California, is the latest escalation in a growing confrontation between the global creative industry and AI companies over the use of copyrighted material in artificial intelligence training.

The music publishers allege that Anthropic obtained and used large quantities of copyrighted musical works without securing the necessary licences from rights holders. They claim the material was incorporated into the development and training of Claude, enabling the AI system to reproduce protected lyrics and generate new content based on material belonging to songwriters and music publishers.

The case could have significant implications for the future of AI development, particularly over whether companies can use copyrighted creative works to train commercial AI systems without compensating their owners.

Music publishers accuse Anthropic of widespread copyright infringement

Sony Music Publishing and Warner Chappell Music allege that Anthropic used hundreds and potentially tens of thousands of copyrighted musical compositions during the development of its AI models.

The complaint reportedly identifies works associated with several globally recognised artists, including The Beatles, Taylor Swift and Michael Jackson, among numerous other musicians and songwriters.

The publishers contend that Anthropic obtained copyrighted lyrics and sheet music through methods including torrent downloads, scraping and other forms of unauthorised acquisition. According to the allegations, these materials were then used as part of datasets that helped train Claude.

The plaintiffs argue that this conduct amounts to copyright infringement because the works were copied and used commercially without permission from the people and organisations that own the relevant rights.

The lawsuit also raises concerns about the scale of the alleged activity. Rather than describing the use of a small number of copyrighted works, the publishers portray the practice as part of a broader effort to collect large amounts of protected creative material for AI development.

Claude allegedly capable of reproducing copyrighted lyrics

One of the central allegations in the lawsuit is that Claude can reproduce copyrighted lyrics when users make specific requests.

The publishers claim that the AI system can generate lyrics that are identical or substantially similar to protected works. They argue that such outputs demonstrate that copyrighted material was not merely used as abstract training information but can remain accessible through the model’s responses.

This has become a particularly sensitive issue for the music industry because song lyrics and musical compositions have significant commercial value.

Music publishers earn revenue through licensing, distribution, performance rights and other uses of copyrighted compositions. If an AI system can provide protected lyrics to users without an appropriate licence, rights holders argue that it could undermine existing markets for authorised access to those works.

The publishers therefore contend that the dispute extends beyond the initial copying of material during AI training. They argue that the resulting capabilities of the AI system may also create direct competition with legitimate copyrighted content.

AI-generated lyrics raise new concerns for songwriters

The lawsuit also focuses on the ability of AI systems to generate new song lyrics.

Sony and Warner argue that Anthropic’s alleged use of copyrighted lyrics helped teach Claude to produce large volumes of AI-generated lyrics. According to the publishers, these outputs could compete with original works created by professional songwriters.

The issue highlights a fundamental concern for the creative sector: whether AI companies should be allowed to learn from copyrighted works and subsequently use that knowledge to produce commercially valuable content without compensating the creators whose works contributed to the system’s development.

For songwriters and publishers, the concern is not simply that an AI system may reproduce an existing song. They are also concerned that models trained on large catalogues of human-created music could generate material that competes with the traditional songwriting market.

Anthropic disputes the allegations

Anthropic has rejected the accusations and indicated that it intends to defend itself in court.

The company has maintained in previous copyright disputes that the use of copyrighted material in AI training can be protected under the fair-use doctrine of US copyright law.

Fair use is a long-established legal principle that can permit certain uses of copyrighted material without permission. However, whether it applies to the large-scale copying of creative works for AI training remains one of the most important unresolved questions in the technology and copyright sectors.

The legal dispute involving Sony and Warner is therefore likely to focus heavily on how existing copyright principles should be applied to modern AI systems.

A growing legal battle over AI training data

The lawsuit is part of a much larger wave of legal action against artificial intelligence companies.

Authors, publishers, artists, musicians and other copyright holders have increasingly challenged technology companies over the use of their work in AI training datasets.

AI models require enormous quantities of information to learn patterns and produce responses. Depending on the system, training datasets can include books, articles, images, computer code, music, lyrics and other forms of creative material.

Copyright holders argue that the commercial use of such material should require permission and compensation.

Technology companies, meanwhile, argue that AI training involves analysing information in ways that can qualify as transformative use and that overly restrictive copyright rules could hinder technological innovation.

The dispute between these two positions is now moving increasingly into courts, where judges are being asked to determine how copyright law should operate in the age of generative AI.

Previous legal challenges add pressure on Anthropic

The latest lawsuit comes after Anthropic faced another major copyright dispute involving authors.

The company previously agreed to a $1.5 billion settlement in a case brought by authors over the use of copyrighted books in AI development. That case became one of the largest financial settlements associated with AI copyright litigation.

The music publishers’ latest action suggests that previous legal settlements have not resolved the industry’s concerns about how AI companies acquire and use copyrighted material.

Sony and Warner are seeking significant financial damages in the new case. Under US copyright law, damages for proven infringement can reach up to $150,000 per copyrighted work in certain circumstances.

Given the potentially large number of works identified in the allegations, the financial exposure could become substantial if the publishers succeed in court.

The plaintiffs are also seeking an order that would prevent Anthropic from continuing to use their copyrighted works in the alleged manner.

Why the case is important for the music industry

The dispute comes at a crucial time for the global music business.

Generative AI is rapidly becoming part of the creative ecosystem. Musicians and producers are experimenting with AI to create melodies, lyrics, arrangements and other forms of musical content.

While some artists see artificial intelligence as a useful creative tool, others are concerned about the possibility of AI systems replicating their work or learning from their creative output without compensation.

Music publishers occupy a central position in this debate because they control or administer rights to large catalogues of musical compositions.

If AI companies are required to obtain licences before using copyrighted music for training, the industry could develop new licensing markets specifically for AI development.

Such a system could potentially provide songwriters and publishers with another source of revenue while allowing technology companies to obtain legally authorised training data.

However, if courts determine that large-scale AI training can generally qualify as fair use, copyright owners could face greater challenges in controlling how their works are used by artificial intelligence systems.

The larger question of copyright and artificial intelligence

At the heart of the dispute is a question that extends far beyond the music industry: how should copyright law apply to AI?

Traditional copyright rules were developed long before machine-learning systems capable of analysing billions of pieces of information existed.

AI companies now process enormous datasets to develop models that can generate text, images, music, software and other forms of content.

This has created a legal grey area concerning whether copying protected works into a training dataset constitutes infringement, whether the resulting AI model represents a transformative use and whether outputs generated by the model can harm the market for the original works.

Courts in the United States are increasingly being asked to address these questions.

The decisions could eventually establish legal standards that affect not only music publishers but also authors, filmmakers, photographers, journalists, software developers and visual artists.

Potential impact on AI companies

A ruling against Anthropic could have consequences for the wider AI industry.

Technology companies may face pressure to establish licensing agreements with copyright owners before using protected material for training. This could significantly increase the cost of developing future AI models.

It could also lead to the creation of new markets for licensed training data, with copyright holders negotiating payments in exchange for allowing AI companies to use their works.

On the other hand, a broad legal victory for Anthropic could give AI developers greater freedom to use copyrighted material for training, although companies could still face restrictions regarding how AI systems reproduce protected works in their outputs.

The outcome could therefore influence the business models of some of the world’s largest technology companies.

What happens next

The lawsuit is expected to become another closely watched case in the rapidly evolving legal battle between artificial intelligence developers and copyright owners.

Sony Music Publishing and Warner Chappell Music are seeking to protect the commercial value of copyrighted compositions and establish greater control over how those works are used in AI systems.

Anthropic, meanwhile, is expected to challenge the allegations and defend its approach to AI training.

The case will ultimately force courts to examine the boundaries between technological innovation and intellectual property protection.

As AI becomes increasingly capable of producing content that resembles human-created work, the question of who should control the data used to train these systems—and whether creators should be compensated—will become increasingly important.

The outcome of the dispute could influence how the AI industry obtains training data, how music publishers license their catalogues and how copyright law is interpreted in the rapidly developing era of generative artificial intelligence.