Sony, EMI and Warner Sue Anthropic Over Music Copyright After $1.5B Book Deal

Three of the world's biggest music publishers, Sony, EMI, and Warner Chappell, filed a copyright lawsuit against Anthropic on Friday, arguing the company's recently settled $1.5 billion payment to authors is nowhere near enough to stop AI labs from pirating training data. The publishers claim Anthropic's illegal torrenting extended beyond books to include thousands of copyrighted musical compositions, and they are pointing to internal staff messages that reportedly celebrate a well-known piracy platform as evidence of deliberate infringement.

What happened

Detail	Fact
Plaintiffs	Sony, EMI, Warner Chappell
Filing date	Friday (most recent)
Prior settlement amount	$1.5 billion (paid to authors)
Books pirated in prior case	More than 7 million
Anthropic valuation cited	$2 trillion
New alleged infringement	Thousands of copyrighted musical compositions

Anthropic, the company behind the Claude AI assistant, recently settled a historic lawsuit brought by authors who proved it had torrented more than 7 million books without permission to use as training data. The $1.5 billion settlement was described at the time as one of the largest AI copyright payouts on record.

Now a group of major music publishers is arguing that settlement set the bar too low. In a lawsuit filed Friday, Sony, EMI, and Warner Chappell contend that “$1.5 billion is obviously not a large enough settlement to deter infringing conduct by a company that has parlayed such mass infringement into a staggering $2-trillion-dollar valuation.” The publishers say Anthropic’s torrenting habit did not stop at books. It also swept up, they claim, thousands upon thousands of their copyrighted musical compositions.

What is the evidence the publishers are citing?

Beyond the scale of the alleged downloads, the publishers are leaning on internal Anthropic staff communications. According to the lawsuit, those chats include messages from employees praising a piracy platform by name, with one message quoted in coverage as “Zlibrary my beloved.” Z-Library is a widely known shadow library site that hosts books and other content without rights-holder permission.

That kind of internal record is significant in a copyright case. It cuts against any argument that the infringement was accidental or that Anthropic was unaware of where its training data came from. Intent matters when courts weigh damages.

Why it matters

This lawsuit is part of a broader pattern: rights holders across every content category, books, music, images, code, are now testing how courts will treat AI training data. The music publishers’ core argument is that the deterrence logic of copyright law breaks down if a $1.5 billion fine is pocket change for a $2 trillion company.

For businesses building on top of AI tools like Claude, the practical risk is indirect but real. If courts agree that training-data infringement was willful and widespread, future licensing costs for AI model providers will rise sharply. That cost eventually flows downstream to API pricing and product fees. It also raises the question of whether outputs from models trained on unlicensed content carry their own legal exposure, a question no court has fully resolved yet.

The music industry has historically been aggressive in copyright enforcement, with established statutory damages that can reach $150,000 per work for willful infringement. If there are truly thousands of compositions at issue, the potential damages figure could dwarf the book settlement.

Our take

The staff chat evidence is the detail worth watching. Proving a corporation “knew” it was infringing is usually the hardest part of a willful-infringement case. Internal messages with explicit praise for a piracy platform hand the plaintiffs exactly that. Anthropic will likely argue that individual employee messages do not reflect company policy, but that defense gets harder when those employees were apparently involved in building the training pipeline.

The $2 trillion valuation framing is deliberate. The publishers are not just seeking compensation; they are trying to set a precedent that settlements must be proportionate to the defendant’s ability to pay and the competitive advantage they gained. If a court buys that argument, it changes the math for every AI lab that built on scraped or torrented data.

For anyone running AI integrations in their business, this is a good moment to ask your vendors exactly where their training data came from and whether they carry any indemnification for copyright claims on model outputs. It is also worth keeping an eye on other active Anthropic litigation, as the company is managing legal exposure on multiple fronts right now.

What to do about it

1. Ask your AI tool providers for their data provenance policy in writing.
2. Check whether your vendor contracts include any IP indemnification clause covering model outputs.
3. Monitor this case: if music publishers win willful-infringement damages, expect price increases across commercial AI APIs within 12 to 18 months.
4. If you produce original creative content, register it. Statutory damages are only available for registered works.

The key takeaway: a $1.5 billion settlement did not close the book on Anthropic’s copyright exposure. It may have opened a new chapter.

Source: Ars Technica · AI