Sony and Warner Sue Anthropic — and Name Both Founders as Defendants
Suing the company is routine. Listing the CEO and a co-founder by name in the caption is not.
Three Key Facts
On Friday, August 28, Sony Music Publishing, Warner Chappell and 35 music-publishing entities in total sued Anthropic in the U.S. District Court for the Northern District of California. The complaint alleges the company obtained tens of thousands of copyrighted musical compositions through torrenting, scraping and bulk downloads from pirate sources, then used them to train the Claude models. The complaint names Library Genesis and Pirate Library Mirror among the pirate sources, and says the downloaded material included books carrying lyrics and sheet music for songs such as Livin On a Prayer, Hallelujah and Great Balls of Fire.
The defendants are not only corporate. Co-founder and CEO Dario Amodei and co-founder Benjamin Mann are named individually. The plaintiffs clearly do not want this filed away as a corporate compliance lapse; they want it pinned to personal decisions. Damages are pleaded at the statutory ceiling: up to $150,000 per work found willfully infringed, plus up to $25,000 for each removal of copyright management information. At the scale of "tens of thousands" of works, the theoretical exposure starts in the billions.
This is the third front in Anthropic's copyright war. In January 2026, Concord and Universal Music Group filed a comparable suit. Before that, the Bartz v. Anthropic authors' case settled at $1.5 billion, and the judge's line from that case keeps getting quoted back: training on copyrighted work is not inherently unlawful, but acquiring it through piracy is. Anthropic says it disagrees with the publishers' claims and intends to defend itself in court.
WangDou's Take
The $1.5 billion Bartz settlement did not buy peace. It published a price list. The judge's distinction — training may be fine, piracy is not — moved the industry's original sin from "how you used the data" to "how you got the data." The second question has server logs, torrent records and purchase invoices attached to it. It is discoverable in a way the first never was.
Which is why the real story here is not the damages number, it is the decision to put Amodei and Mann in the caption. The plaintiffs are betting that the early "grab the corpus now, sort it out later" calls were made by identifiable people. If a court accepts that framing, five years of industry-wide silence about data provenance stops being a shared understanding and starts being one executive's email.
The dividing line for this quarter is drawn. Companies that can document clean data sourcing get a moat priced into their valuation. Companies that cannot need a litigation reserve on the balance sheet first.
Source: TechCrunch, Axios