Authors Sue Major AI Companies Over Alleged Use of Pirated Books for Training
Six authors have filed new individual copyright infringement suits in the Northern District of California against Anthropic, OpenAI, Google, Meta, xAI, and Perplexity. The filings claim the companies copied books from pirate libraries-LibGen, Z-Library, and OceanofPDF-to train their models without permission or payment.
This group opted out of the proposed $1.5 billion settlement connected to an earlier case against Anthropic. They argue that high-quality books are "gold standard" training data and that relying on pirated copies to build commercial systems deserves far more than the $3,000 per work proposed in that settlement.
What's new-and why it matters for working writers
- The suits focus squarely on alleged use of pirated ebooks, the same factor a previous judge highlighted as requiring a remedy.
- Authors say $3,000 per work is too low-about 2% of the statutory maximum for willful infringement.
- They're seeking $150,000 per work per defendant-$900,000 per work across the six companies.
- These are individual actions, not a class action; each author seeks a separate jury trial.
- The attorneys are fronting costs in exchange for a 35% contingency fee.
- Opt-out deadline: Authors have until January 16 to opt out of the Anthropic settlement.
- ClaimsHero, a site encouraging opt-outs, was criticized in court for misleading communications and ordered to make changes.
- Company responses so far are sparse; xAI replied "Legacy Media Lies." Perplexity's communications lead said the company does not index books.
- No preliminary hearing date yet.
The legal angle in plain English
A prior ruling found AI training can qualify as fair use-but flagged the use of pirated books as a separate problem that warrants compensation. That's why these new filings center on the sources of the training data rather than the concept of model training itself.
Statutory damages for willful infringement can reach $150,000 per work. The authors argue the $3,000 proposal is far below that ceiling. You can review the statute here: 17 U.S.C. § 504 (statutory damages).
Who's suing and who's named
- Plaintiffs: John Carreyrou; Lisa Barretta; Philip Shishkin; Jane Adams; Matthew Sack; Michael Kochin.
- Defendants: Anthropic, OpenAI, Google, Meta, xAI, Perplexity.
The filing notes the authors' backgrounds range from investigative nonfiction and psychology to technology, politics, and spirituality. Two outside firms-Stris & Maher LLP and Freedman Normand Friedland LLP-are listed on the suits.
What writers should do right now
- Register your works. It strengthens your position if you ever need to enforce your rights. Start here: U.S. Copyright Office Registration.
- Audit your catalog. Search for your titles on known pirate libraries and document what you find (screenshots, URLs, timestamps).
- Decide on settlement vs. litigation. If you're part of the proposed Anthropic settlement class, calendar the January 16 opt-out date and talk to a lawyer before acting.
- Read the fine print. Contingency arrangements (like 35%) and cost advancement terms matter more than the headlines.
- Preserve evidence. Save contracts, editions, registration certificates, and proof of publication dates.
- Update your boilerplate. Add clauses in new publishing and licensing deals addressing AI training and dataset use.
- Monitor model outputs. Keep examples where an AI system appears to reproduce identifiable passages from your work-note time, prompts, and results.
Context and next steps
These cases test how courts weigh alleged use of illicit sources against broad claims of fair use for AI training. They also test whether individual suits-rather than class actions-can force higher payouts per work.
Expect months of motions before anything reaches a jury. If you're considering action, get counsel early, collect documentation now, and set reminders for the opt-out window if it applies to you.
Further learning
- Want a structured overview of current AI tools (so you can set clearer licensing boundaries)? See this curated list: AI courses by job.
Your membership also unlocks:
