A coalition of music publishers has filed a federal lawsuit against Anthropic in Northern California. The complaint alleges the AI developer used copyrighted song lyrics to train its Claude model without authorization.
The $150,000-per-song demand
The lawsuit targets not only Anthropic as a corporate entity but also its co-founders, Dario Amodei and Benjamin Mann. According to the report, the plaintiffs are seeking $150,000 for every single instance of unauthorized use of copyrighted lyrics. Because the suit involves thousands of tracks, the total financial liability for Anthropic could potentially reach into the billions of dollars .
This aggressive pricing strategy suggests that the music industry is no longer interested in mere apologies or small settlements. By naming the founders personally, the publishers are attempting to create maximum pressure on the leadership of the AI firm to settle the dispute quickly.
From 'Eye of the Tiger' to 'Paper Rings'
The core of the infringement claim rests on the training data used for the Claude language model. As the report says, the plaintiffs allege that Anthropic sourced data from copyrighted lyric collections as well as public repositories like LyricFind and MusixMatch.. The filing provides specific examples of songs that the model allegedly reproduces in verbatim or highly recognizable fragments, including "Eye of the Tiger," "Paper Rings," "All I Want for Christmas Is You," "Ain't No Mountain High Enough," and "Here Comes Santa Claus."
Copyright holders argue that when an AI model like Claude can output these lyrics, it directly undermines the commercial value of licensed content. This creates a scenario where the AI becomes a substitute for the licensed product, stripping publishers of their exclusive rights to distribute and monetize the written work of songwriters.
The $1.5 billion shadow of Batz v. Anthropic
The current litigation is not the first time Anthropic has faced accusations of data theft. The complaint draws a direct parallel to the 2022 Batz v. Anthropic case, which resulted in a $1.5 billion settlement. In that previous instance, the AI firm was found to have used pirated academic materials and books harvested from Torrent platforms, specifically citing sources like Pirate Library Mirror and Library Genesis.
The music publishers are using the Batz case to establish a "pattern of disregard" for intellectual property law. By highlighting that Anthropic was previously forced to delete illicit copies of books, the plaintiffs argue that the company is a repeat offender that views copyright infringement as a standard part of its development process.
The missing names of the coalition publishers
Despite the scale of the lawsuit, several critical details remain unverified. The source mentions a "coalition of music publishers" but does not name the specific companies or individuals leading the charge. Furthermore, while the suit mentions "thousands of tracks," the exact number of songs being claimed as infringed has not been disclosed.
It is also worth noting that the report does not include a response from Anthropic, Dario Amodei, or Benjamin Mann. Whether the company will argue "fair use"—a common defense in AI training cases—or attempt to settle out of court remains to be seen.
The risk of mandatory training data audits
The outcome of this case could fundamentally change how AI models are built. If the Northern California court rules in favor of the publishers,AI developers may be forced to conduct exhaustive audits of their training sets and secure expensive licenses for every piece of data used. This would significantly increase the cost of development and potentially slow the release of new iterations of Claude and other competing models.
Conversely, a victory for Anthropic would likely embolden other AI firms to continue aggressive data-scraping strategies. Such a ruling would signal to Fortune 500 publishers and independent creators alike that their intellectual property is effectively fair game for the purpose of advancing large-scale language technology.
Comments 0