Major publishers and author Scott Turow struck hard Tuesday. They filed a class-action lawsuit in Manhattan federal court against Meta Platforms and its CEO Mark Zuckerberg. The core charge: willful copyright infringement on a staggering scale. Meta allegedly torrented over 267 terabytes of pirated books, journals, and articles from sites like LibGen, Sci-Hub, and Anna’s Archive. All to train its Llama AI models.
Plaintiffs include Elsevier, Cengage, Hachette Book Group, Macmillan Publishers, and McGraw Hill. Turow, whose novels like Presumed Innocent have sold 40 million copies, joins them individually. Their complaint paints a picture of deliberate theft. “Defendants Meta and Zuckerberg followed their well-known motto: ‘move fast and break things,’” it states. They illegally downloaded millions of protected works. Then copied them repeatedly for Llama training. This, plaintiffs say, ranks as one of history’s largest copyright violations. (Variety; Association of American Publishers)
Internal Meta documents fuel the fire. A December 2023 employee memo flagged LibGen as a pirated trove. Legal risks? High. Disclosure? None planned. Yet executives pushed ahead. Zuckerberg himself looms large. As founder, chairman, and controlling shareholder, he “personally authorized and actively encouraged” the effort, the suit claims. One email chain shows licensing talks halted after escalating to him. A staffer noted: buying even one book would undermine fair-use arguments. Meta had budgeted up to $200 million for deals. That vanished in April 2023.
From Pirate Torrents to AI Outputs
And the scale stuns. 267 TB equals hundreds of millions of publications. Far bigger than the Library of Congress print holdings. Meta stripped copyright notices from files. Hid sources. Llama now spits out substitutes: verbatim excerpts, fake chapters, knockoff novels mimicking Turow’s style. “Llama even tailors outputs to mimic the expressive elements and creative choices of specific authors,” the complaint alleges. Users prompt it for textbooks, papers, stories. It delivers—built on stolen inputs.
Publishers argue this kills markets. Why license when AI generates free alternatives? Meta signed deals with news outlets like CNN and Fox. African publishers too, for limited sets. But books and journals? Targeted for their structure—length, coherence, editing. Perfect LLM fodder. Harm hits at ingestion, they say. No consent. No pay. Just copies fueling a multibillion-dollar system.
Meta fights back. A spokesperson told Variety: “AI is powering transformative innovations… courts have rightly found that training AI on copyrighted material can qualify as fair use. We will fight this lawsuit aggressively.” Fair use. The battle cry for OpenAI, Anthropic, Google too. But plaintiffs counter: this isn’t passive scraping. It’s torrenting known pirates. Circumventing protections. Abandoning licenses on Zuckerberg’s word.
Turow knows courts. He’s chaired the Authors Guild. His books fuel Hollywood. Now he’s plaintiff. The suit seeks cash damages. An injunction. Destruction of infringing copies. Class certification for other owners. Lawyers from Oppenheim Zebrak, Debevoise & Plimpton, and Keller Rohrback lead. “Built a multibillion-dollar empire on the backs of publishers and authors,” they charge.
Past Rulings Haunt, But Gaps Emerge
Precedent cuts both ways. Last June, a judge tossed authors’ claims against Meta—including Sarah Silverman’s. Training Llama 1? Fair use. No direct copying alleged then. This suit pivots. Willful acts. Pirate focus. Zuckerberg’s hand. Earlier amicus briefs slammed Meta’s “shady” datasets. (Missouri Lawyers Media)
So where does this land? Federal court in New York. Southern District. Judges there eye AI cases closely. NYT beat OpenAI on discovery. Anthropic faces similar heat. Canada’s media sued OpenAI last December. (CyberBeat News) Disney hit Midjourney, then Meta bots. Hollywood fumes. But books? Publishers hold copyrights tight. Llama’s open weights spread the model wide. Outputs flood markets.
Short-term: motions fly. Discovery digs emails, memos. Long-term: fair use tests limits. Does “move fast” trump ownership? Billions ride on it. Meta’s stock holds steady. Publishers? They’ve waited years. Turow’s thrillers outsell most. Now reality bites harder.
Zuckerberg built Facebook on photos, posts. Llama eyes the world’s knowledge. Pirates supplied. Courts decide if that’s theft—or progress.
WebProNews is an iEntry Publication