Legal Battle Escalates: Pulitzer Prize-winning author John Carreyrou and other writers file a copyright infringement lawsuit against OpenAI, Anthropic, Google, Meta, xAI, and Perplexity, accusing them of stealing their literary works from illicitly sourced book repositories. This marks the first legal action targeting both xAI and Perplexity over alleged AI training violations.
Six major artificial intelligence companies are now facing a fresh wave of litigation. According to Bloomberg Law, a group of authors—including Pulitzer Prize-winning journalist John Carreyrou—filed suit this week in the U.S. District Court for the Northern District of California.
The complaint names Anthropic, Google, OpenAI, Meta, xAI, and Perplexity as defendants, alleging "willful piracy" by these tech giants. Notably, this is the first copyright case directed at xAI’s model training practices and the initial lawsuit brought by authors against Perplexity.
The filing asserts that these AI firms obtained copies of books from so-called “shadow libraries,” including notorious piracy platforms such as LibGen, Z-Library, and OceanofPDF.
Plaintiffs argue that the defendants violated copyright law twice: first by illegally downloading protected works, and again by generating additional unauthorized copies during model training or optimization processes. The authors claim their creations have become “foundational components” of multi-billion-dollar AI product ecosystems—yet they received no compensation whatsoever.
xAI Dismisses Claims as 'Legacy Media Fiction'
As reported by Bloomberg Law, xAI responded to the allegations with a brief statement labeling them as “legacy media fiction.” Meanwhile, Perplexity spokesperson Jesse Dwyer stated the company does not index books. However, the complaint contends that Perplexity has reproduced and exploited copyrighted content within its retrieval-augmented generation (RAG)-powered AI search system without authorization.
A recent similar case concluded with Anthropic agreeing to a $1.5 billion settlement with a group of authors. That ruling established a key legal precedent: even if AI training could qualify under fair use doctrine, it cannot shield companies that relied on pirated material.
Carreyrou and several co-plaintiffs opted out of joining the class-action settlement against Anthropic, arguing that the proposed payout—approximately $3,000 per work—is negligible compared to actual market value. By pursuing individual litigation instead, the new case may result in significantly higher damages. If a jury finds the companies engaged in intentional infringement, plaintiffs could receive up to $150,000 in statutory damages per infringed work.
The complaint emphasizes: “Large language model developers must not be allowed to trivialize thousands of high-value claims with minimal settlements, thereby evading the true cost of their widespread and deliberate acts of copyright violation.”
