Meta AI Copyright Lawsuits: 292 Authors Skip Class Action to Sue

A new legal strategy against Meta

Meta Platforms is facing a new wave of copyright lawsuits over how it built its AI models. This time the authors are suing one by one instead of joining a group case. On Monday, October 19, attorneys from Freedman Normand Friedland and Stris & Maher filed claims against Meta in the U.S. District Court for the Northern District of California. They allege that Meta copied the authors' books without permission and used them to train its large language models.11 Every author is pursuing an individual claim and asking for statutory damages, which the Copyright Act caps at $150,000 per infringed work.11

The exact number of plaintiffs depends on when each outlet counted. Bloomberg Law put it at roughly 300 authors.11 Publishers Lunch reported 231 authors across four complaints, led by Jacqueline Winspear, Daniel José Older, Robert Sheard and Tosca Reno.17 A docket tracker later found a fifth near-identical complaint, Levine v. Meta, which brought the total to 292 authors, with 56 to 61 plaintiffs in each filing.8 Publishers Weekly also counted five complaints and said the authors were split among them mostly for logistical reasons.12 The 292 figure appears to be the most complete count. The gaps between outlets come from a filing spree that kept going after the first reports were written.

The plaintiffs range from well-known novelists to journalists and academics. They include Winspear, author of the Maisie Dobbs mysteries; memoirist Reyna Grande; journalist Rachel Nuwer; and China scholar Minxin Pei.12

Why the authors chose not to join a class action

The filings break with how most AI copyright cases have been run. Plaintiffs' firms have usually filed class actions, where a few named plaintiffs represent thousands of rights holders.11 The complaints criticize that model openly. They say the authors want full control of their own cases instead of being folded into large settlements that resolve claims "for pennies on the dollar."17

The main target of that criticism is the Anthropic settlement. The first authors' class action against Anthropic ended in a $1.5 billion deal that pays about $3,000 per work.1112 The new plaintiffs want juries, not judges, to set damages, and they are aiming close to the statutory maximum for willful infringement.12 Kyle Roche of Freedman Normand Friedland told Publishers Weekly that his firm represents more than 2,200 authors. He said he plans to file up to 1,000 individual claims by the end of the year.12 The firms told Bloomberg Law they would sue for close to 1,000 more authors in the coming months.11

The client recruitment also stands out. All of the authors signed up through ClaimsHero, a platform that partnered with the two firms last fall to persuade authors to opt out of the Anthropic settlement and sue on their own.12 A blog that tracks AI copyright cases compared the approach to mass tort litigation. It asked two practical questions: why file five nearly identical complaints instead of one, and whether all 1,000 plaintiffs will sit for depositions.18

The Authors Guild gave a mixed response. CEO Mary Rasenberger backed the authors and praised the complaints for treating Meta's book scanning as a separate act of infringement. She also said authors should stick together through well-designed class actions.12 That tension matters. If many authors opt out to chase bigger individual awards, future class settlements could shrink and get harder to negotiate. This new route could make collective bargaining harder for the authors who don't take it.

What the complaints allege, including Project Vellum

The core allegations are familiar. The complaints say Meta downloaded hundreds of thousands of e-books from shadow libraries, which are pirate sites.11 Winspear names LibGen, Z-Library, Anna's Archive and the Books3 dataset. It says Meta often used BitTorrent, a file-sharing method that also passes copies to other users, and that Meta removed copyright notices from the files.12

The new element is a claim about Project Vellum. According to the complaint, Meta began this program around mid-2025, after it had already been sued over Llama. Meta allegedly bought physical books in bulk, scanned them, and used the text to train its Muse Spark and Muse Glimmer models, which came out in April and August 2026.12 The authors argue that buying a book gives you ownership of that one copy, not the right to reproduce it.12 This claim is aimed at the line Judge William Alsup drew in the Anthropic case. He found AI training to be fair use but said the use of pirated books required a remedy.12 If the authors can show that even lawfully bought books cannot be scanned for training without a license, the exposure would extend beyond piracy to legitimate data pipelines.

Court rulings so far

In court, the authors are working against precedent. In Kadrey v. Meta, Judge Vince Chhabria ruled that Meta's use of the books for training was fair use. He found the plaintiffs had not shown harm to the market for their works.1112 He also said he might have sided with the authors if they had proven market dilution, meaning that a flood of AI-generated books undercuts sales of human-written ones.1210 That argument will likely be central to the new cases. Plaintiffs' filings on it are due December 16.12

Chhabria now oversees all the related Meta cases.12 The new suits will probably be assigned to him as part of the Kadrey group.18 Bloomberg Law also reported in July that Meta's generative AI copyright ruling failed to get a quick appeal.11 Separate cases from Chicken Soup for the Soul, academic publisher Cognella, and journalist John Carreyrou's company Cambronne are set for jury trials starting May 3, 2027.12

Meta also faces a larger proposed class action filed in May in Manhattan by Elsevier, Cengage, Hachette, Macmillan, McGraw Hill and author Scott Turow.16 That complaint names Zuckerberg personally.10 It alleges Meta considered spending up to $200 million on licensing before switching to pirated sources, and it quotes an internal message: "If we license one single book, we won't be able to lean into the fair use strategy."3 Meta has said courts "have rightly found" that training on copyrighted material can be fair use, and it has promised to fight aggressively.39

Discovery is moving toward Meta's leadership

The more important ruling may have come the same day the individual suits were filed. On October 19, Magistrate Judge Thomas Hixson ordered Meta to identify every person who decided on or approved getting data from shadow libraries and using BitTorrent. Where decisions were made by teams, he said every team member must be named.16 Meta had argued that its early dataset work was informal and that no one with authority made the decision. Hixson rejected that.16 The order also requires command-history files for the servers Meta used to torrent.15 Plaintiffs say those files could show whether Meta seeded, or re-shared, the files it downloaded. Meta says it took steps to prevent seeding.1516

The numbers are large. Plaintiffs allege Meta pulled about 80.6 terabytes from LibGen and 35.7 terabytes from Anna's Archive. They also point to internal messages referring to "prior escalation to MZ" before the LibGen download.16 Separately, Chhabria refused in August to block a deposition of Zuckerberg, and Roche then deposed him over Meta's objections.12

This is where the cases converge. Fair use has mostly protected the act of training. Willfulness is what pushes statutory damages toward $150,000 per work, and willfulness depends on who knew what. Each discovery order that moves up Meta's chain of command strengthens the individual plaintiffs' case for high jury awards. That is the bet behind the opt-out strategy.

What this means for AI developers and the legal industry

For AI developers, the lesson is about data provenance. Legal analysts quoted in coverage of the Hixson order expect AI labs to rework how they acquire data and to require strict provenance audits for training sets.15 The order also warns companies that informal approval processes can work against them in court.16

For the legal industry, the story is less about AI tools in law firms than about how AI litigation itself is being organized. Lawyers are using an online platform to sign up thousands of authors, then filing near-identical complaints in batches. That takes legal claims that have mostly been handled by class actions and runs them through a high-volume process that looks like mass torts.1218 The number of copyright suits against AI companies recently passed 150, according to one tracker.18 If juries in 2027 award damages well above the Anthropic benchmark, expect other firms to copy this model. If Chhabria's fair-use framework holds and market dilution goes unproven, these cases may end up confirming that the class settlement was the better deal.

My reading: the opt-out strategy is a reasonable bet, but it relies on willfulness, and right now the evidence on that question is going against Meta.