On a quiet August afternoon in a San Francisco federal court, two of the world's most powerful music companies drew a line in the sand. Sony Music and Warner Chappell have filed suit against Anthropic in the US District Court for the Northern District of California, alleging the unauthorised use of "tens of thousands" of copyrighted works to train the company's Claude series of artificial intelligence models. At the maximum statutory rate of $150,000 per work — plus an additional $25,000 for each instance in which copyright data was allegedly stripped from a file — the total damages sought could reach several billion dollars. What makes this complaint unusually sharp is not merely its scale, but its specificity.

The suit alleges that co-founder Benjamin Mann personally used BitTorrent to download more than five million pirated books, while other Anthropic employees are said to have acquired at least two million additional titles from the Pirate Library Mirror. The complaint further claims that Anthropic scraped song lyrics from platforms such as MusixMatch and LyricFind — services that had themselves paid licensing fees to the very labels now suing. Among the specific tracks identified in the training data are Leonard Cohen's "Hallelujah," Earth, Wind & Fire's "September," and Taylor Swift's "Paper Rings" — a catalogue that spans continents, generations, and genres. This lawsuit does not arrive in isolation.

Anthropic recently settled a case brought by the publishing industry for $1.5 billion, and has faced separate legal action from Universal Music Group, Concord, ABKCO, BMG, and Round Hill Music. The pattern suggests that the creative industries, from Lagos to London, are no longer willing to treat AI training data as a legal grey area.