35 Music Publishers Sued Anthropic on August 28 and Named Dario Amodei Personally. If Your Product Runs on Claude, Read the Complaint.

On August 28, thirty-five music publishers, including Sony Music Publishing and Warner Chappell, filed a 48-page copyright complaint against Anthropic in the Northern District of California. They named CEO Dario Amodei and co-founder Benjamin Mann as individual defendants, not just the company. They're seeking up to $150,000 per infringed song across "tens of thousands" of compositions, plus $25,000 for every instance of stripped copyright management information. If Anthropic loses even a fraction of that exposure, the number lands somewhere in the billions.

I build things on top of Claude. So do a lot of people reading this. The interesting part of this suit isn't whether the publishers are right (I have no idea, and neither do you yet). It's what "your AI vendor is being sued for billions" actually means for the person one layer up the stack.

What the complaint alleges

The publishers' claim is specific: they say Anthropic acquired copyrighted song lyrics through "torrenting, scraping, and downloading" copyrighted works at scale, then used those copies to train Claude. The complaint lists specific songs, including "Ain't No Mountain High Enough," "All I Want for Christmas Is You," and "Eye of the Tiger." It's not a vague "AI ate my work" claim, it's a list of specific compositions the publishers say they can trace into Anthropic's training pipeline, and a specific method of acquisition they say was unlawful.

That last part matters more than it looks like it does, because it's exactly the question a different court already ruled on.

Why this isn't the same case Anthropic already won

Anthropic has been here before, and it didn't go the same way for every claim. In the Bartz book case, Judge William Alsup found that training on books Anthropic had legitimately purchased was fair use, even when the physical copies were destructively scanned afterward. But he denied Anthropic summary judgment on books acquired through pirate sources like LibGen, and that half of the case settled for $1.5 billion covering roughly 500,000 works, the largest copyright settlement in US history.

The line that case drew wasn't about training being harmful. It was about how the copy got acquired. Buy it, and format-shift it, and you're fine. Torrent it, and you have a problem. The music publishers' complaint is built to land on the wrong side of that exact line: it explicitly alleges torrenting and scraping, not purchase. If the fact pattern holds up the way the complaint describes it, Anthropic is arguing from a weaker position than it had with the books, not a stronger one.

Why naming the founders personally is the signal worth noticing

Suing a company is normal. Naming a CEO and co-founder as individual defendants in a corporate copyright case is a more aggressive move, and it usually means the plaintiffs are trying to either pierce a shield they think is thin or apply pressure that a corporate defendant alone doesn't feel. Whether that claim survives a motion to dismiss is a separate question from whether it works as a signal to the market: this is now a fight the publishers plan to make personal, not procedural.

I'm not a lawyer, and nothing here is legal advice. But you don't need a law degree to read "we are suing your CEO by name" as a statement of intent about how far this is going to go.

What "vendor legal risk" actually means for you

Here's the part that's easy to skip past if you don't build on foundation models for a living. If you ship a product that's a thin layer over the Claude API, calling Claude API is a supplier relationship, and this lawsuit is a live demonstration of what supplier risk looks like when the supplier is a foundation model company mid-lawsuit.

A worst-case outcome for Anthropic here isn't hypothetical: an injunction, a forced retraining, or a settlement that changes API terms or pricing overnight are all things that have happened to AI labs in the last two years, just not usually to Anthropic specifically and not usually over music. None of that is something you did wrong. You didn't scrape anything. You just built on a platform, the same way you'd build on AWS or Stripe. But unlike AWS or Stripe, this particular category of vendor is currently fighting existential-scale copyright litigation on a regular cadence, and you have approximately zero visibility into how it resolves.

What I'd actually do

I'm not telling you to stop using Claude, and I haven't stopped either. What I have started doing is treating model-provider concentration the way I'd treat any other single point of failure in a business: write down what breaks if this vendor is unavailable, rate-limited, or forced into a pricing change for thirty days, and have an actual answer, not a shrug. For most of us that means keeping at least one other model's API wired into the codebase even if you never route traffic to it under normal conditions, so that switching is a config change instead of a rewrite under pressure.

The honest counter to my own argument: this suit was filed a week ago, discovery hasn't started, and most of these cases either settle quietly or get whittled down well before they reach a number that changes anyone's day-to-day API access. I could be reacting to a headline that turns into a footnote. But the cost of a backup integration is a weekend, and the cost of not having one during the thirty days you actually need it is your whole product being down. That asymmetry is why I'd rather over-prepare for a low-probability event than assume my vendor's legal department has it handled.

Author

Lukas

@lukcombinator

Sources

• Sony Music, Warner sue Anthropic, alleging a 'brazen campaign' of intellectual property theft (TechCrunch)
• Sony, Warner Sue Anthropic for Allegedly Illegally Training Claude (Variety)
• Anthropic copyright lawsuit brought by 35 music publishers (Digwatch)
• What Authors Need to Know About the $1.5 Billion Anthropic Settlement (Authors Guild)