What the 300 authors allege

Nearly 300 authors filed copyright infringement claims against Meta on October 19 in the U.S. District Court for the Northern District of California. They allege Meta trained its AI models on pirated copies of their books and, more recently, on physical books the company bought and scanned without a license, according to Publishers Weekly.

The exact headcount depends on which filing count you trust. ChatGPT Is Eating the World reported 292 authors spread across five near-identical suits, with individual cases carrying roughly 56 to 61 authors each. One suit, filed by Jacqueline Winspear and about 60 co-plaintiffs, also targets Meta's newer Muse models alongside Llama. Treat the total as approximate until the court records settle it.

A docket entry on Justia confirms one of the filings, Reno et al. v. Meta Platforms, Inc., Case No. 3:2026cv12131. The complaint was filed October 19, 2026, with a jury demand from the plaintiffs. These cases are at an early stage, and no court has ruled on their merits.

Why individual claims change the math for AI companies

The authors chose individual claims instead of joining a class action. They want a jury trial and the maximum remedy available for each act of infringement. According to FourWeekMBA, statutory damages run from $750 to $30,000 per work and can rise to $150,000 for willful infringement.

Per-work damages scale with the size of a catalog. A company that copied thousands of titles faces exposure that grows with each one. A class action can resolve claims through a single negotiated pool of money. Individual suits keep each claim separate, and each one carries its own potential award.

An Authors Guild spokesperson, quoted by Publishers Weekly, argued that authors “must stick together” through well-conceived class actions to protect all writers. That position runs opposite to the path these 300 authors chose.

The settlement benchmark is Bartz v. Anthropic, which ended with a $1.5 billion payment. Tech-Insider reports that one source puts the figure at roughly $3,000 per work, and Flowmatix says the settlement received final approval in July 2026. Meta's lawyers will weigh that number, and so will the authors.

A judge orders Meta to name who approved the downloads

The same day, U.S. Magistrate Judge Thomas Hixson ordered Meta to identify each person who made, authorized or approved decisions to obtain or download shadow-library datasets containing the plaintiffs' works. The order arises in the consolidated Kadrey v. Meta case and in a separate suit brought by Entrepreneur Media, according to BigGo Finance and Cryptobriefing.

The order does not decide whether Meta infringed copyright, and it does not decide whether any executive authorized piracy. Geekslop describes it as a discovery ruling that sets what Meta must disclose while the litigation continues. That distinction matters, because the answers will come from discovery rather than from a verdict.

The plaintiffs' allegations are specific. They claim Meta obtained roughly 80.6 terabytes of data from LibGen and about 35.7 terabytes from Anna's Archive. They also say Meta answered “no one decided” even though approval was required, according to BigGo Finance.

That answer gives the court a reason to press for names. If the honest response to “who approved this” is no one, the next question is who had the authority to say yes. For Meta, the exposure now reaches beyond engineers. Discovery could expose executives, and it could change how AI companies approve training data inside their own walls.

Project Vellum and the move to physical books

According to the complaint as reported by Publishers Weekly, Meta ran a book-scanning program called Project Vellum starting around mid-2025. That was after the company had already been sued over Llama. The scanning adds a second acquisition theory to the piracy claims, and it draws on books Meta paid for in physical form.

The legal question shifts with physical books. The only federal court to consider physical-book scanning so far, Judge Alsup in Bartz v. Anthropic, held that scanning was fair use. The same ruling rejected fair use for copies obtained from shadow libraries. Whether Project Vellum scanning qualifies as fair use under Ninth Circuit law is still unresolved, according to ChatGPT Is Eating the World.

That split explains the authors' focus on acquisition. A book that was scanned from a lawfully purchased copy presents one argument. A copy pulled from LibGen or Anna's Archive presents another. For a sense of how much Meta is spending on the models at issue, see our report on Meta lifting its 2026 AI capex floor.

How the Kadrey ruling shaped the new strategy

Judge Chhabria, who presides over Kadrey v. Meta, ruled in June 2025 that copying the named plaintiffs' books to train Meta's models was fair use on the record before him. Claims tied to torrenting remained unresolved. He also said he might have sided with the authors had they shown that training harmed the book market through dilution, per Publishers Weekly.

The new complaints read like a response to that ruling. Instead of arguing only that training is unlawful, they focus on how the books arrived and on who inside Meta made the call. Meta's earlier win on training gives the plaintiffs a reason to shift their argument toward the source of the data.

The wider docket is large. One tracker counts 149 copyright lawsuits against AI companies in the United States, according to ChatGPT Is Eating the World. A Third Circuit affirmance on September 29, 2026 upheld the Thomson Reuters v. Ross ruling that AI training competing with the source was not fair use, according to the AI Lawsuit Tracker. Ross involves different facts from Meta's cases, so it does not settle the questions here.

Publishers filed a rival class action in May

Big publishers took the class route first. In May 2026, Hachette, Macmillan, McGraw Hill, Elsevier and Cengage, along with novelist Scott Turow, filed a proposed class action in Manhattan federal court against Meta and CEO Mark Zuckerberg. The complaint centers on alleged shadow-library sourcing, according to BigGo Finance.

Put the two tracks side by side. The publishers pursue one class-wide case that could settle for a pooled sum. The authors pursue separate suits with per-work damages. Meta could face both approaches at once, and a court may treat them differently.

Meta has not been quoted in the sources reviewed on the October 19 complaints, so its position remains unknown.

What authors and readers can do now

If you are an author who has been contacted about joining one of these suits, take a few concrete steps before you sign anything:

  • Read every fee, contingency and settlement term in the agreement.
  • Ask whether the suit is a class action or an individual claim, and what happens to your claim if the case settles.
  • Talk to a copyright attorney before you sign.
  • Check the docket. Case No. 3:2026cv12131 is searchable on Justia and in the federal court system.

If you are a writer or rights holder who thinks your work was used, you do not need to act immediately. Watch for class notices, and check whether you are a plaintiff in any of these cases.

What to watch next

Three developments will shape the next phase. The first is whether Judge Chhabria takes the new cases, which one source expects but has not confirmed. The second is the list of names Meta provides under the Hixson order, which will show whether any executive approved torrenting. The third is whether the Ninth Circuit addresses Project Vellum scanning directly.

No court has yet ruled on the merits of the October 19 complaints. Until a judge does, the most concrete evidence will come from the discovery record. If that record points to senior decision makers, the dispute moves from training data toward corporate governance. That shift could push other AI companies to audit where their book collections came from before the next lawsuit arrives.