Sony Music Publishing and Warner Chappell Music filed a multibillion-dollar copyright infringement suit in the Northern District of California against Anthropic, accusing the company of illegally scraping and downloading copyrighted works to train Claude AI models.

The complaint cites specific songs including Eye of the Tiger, Ain’t No Mountain High Enough, All I Want for Christmas Is You, and Paper Rings, with alleged sources such as Library Genesis and Pirate Library Mirror.

Damages sought could reach up to $150,000 per work, in addition to other remedies requested by the plaintiffs.

For end users and the wider market, the ruling could influence debates on content originality, compliance tools, and how AI systems are designed around legally sourced training data.

A key legal distinction in this case is its focus on data acquisition (torrenting) rather than the training mechanics, reframing the question as the legality of acquiring training data.

The report frames ongoing fair-use debates in AI training and notes a separate high-profile case involving the New York Times, underscoring continued uncertainty around training-data rights.

Implications include: audits of training corpora for frontier labs, alignment of IPO timelines with ongoing litigation creating contingent liabilities, and a shift in AI legal risk from training methods to data acquisition practices.

The case highlights broader industry challenges around data provenance and copyright in AI, pressuring developers to secure licenses, use licensed/public-domain content, or improve data filtering and synthetic data strategies.

If the acquisition theory wins in discovery, it could establish training-data provenance as a primary legal exposure for AI labs, potentially requiring data-acquisition logs and scrutiny of founders’ decisions.

The central questions concern whether AI training can be fair use and whether acquisition methods undermine copyright protections, with wide implications for data pipelines used by AI labs.

The case targets the data acquisition pipeline itself, potentially setting a precedent that how data is obtained can be a standalone copyright violation in AI training.

The new case sits in the Northern District of California, drawing on related Bartz v. Anthropic litigation and placing Anthropic under renewed scrutiny over data sourcing and training practices.