A federal judge has drawn the first durable line in AI copyright law — and it runs right through Stanford's English department. Judge William Alsup ruled that Anthropic illegally scraped copyrighted books from shadow libraries like LibGen and PiLiMi to train its Claude models, then blessed the legal acquisition of books for AI training as fair use. The distinction matters: piracy is out, but mass ingestion of purchased or licensed works is now judicially endorsed. The $1.5 billion Bartz v. Anthropic settlement covers nearly 500,000 registered works, yielding roughly $3,000 per infringed work. Tobias Wolff, professor emeritus of English and National Medal of Arts winner, calls this inadequate. Ten of his publications were used without permission. He and Chang-rae Lee have opted out, filing Cruz v. Anthropic seeking up to $150,000 per work in statutory damages — a 50x multiplier over the class settlement. They want a permanent injunction against Claude training on their works. Alsup's fair use reasoning reveals the judicial frame taking shape: AI training is analogized to human reading. "Like any reader aspiring to be a writer, Anthropic's LLMs trained upon works not to race ahead and replicate or supplant them — but to turn a hard corner and create something different," he wrote. This is the metaphor that will structure the next decade of AI copyright doctrine. It elides the industrial scale, the commercial motive, and the asymmetry between a reader who buys a book and a company that ingests 500,000 of them to build a product worth tens of billions. The structural economics are stark. Anthropic captures the generative capacity of half a million works and converts it into commercial AI products. Authors receive a one-time payout that amounts to less than a decent advance on a single book. Stanford IP scholar Paul Goldstein identifies the core problem as transaction costs — it is simply impracticable for AI companies to license millions of individual works. This framing conveniently positions the status quo as inevitable rather than chosen. Over 70 infringement lawsuits are now filed against AI companies, including actions by The New York Times and Getty Images. The Authors Guild is pursuing OpenAI and Microsoft separately and lobbying Congress for mandatory licensing legislation. The legal infrastructure is fragmenting across class actions, individual suits, and legislative tracks, with no single resolution in sight. Wolff frames the stakes generationally: established writers with careers and backlists will survive. Young writers building careers face an economy where their work can be absorbed at industrial scale for pennies. "To put me to work that way, without remittance, is a form of slave labor," he said. The U.S. remains the only industrialized nation without a moral right of attribution for authors — a gap that AI training has turned from an academic concern into an economic one. The Bartz ruling sets a floor, not a ceiling. Piracy is banned; mass ingestion is blessed. The extraction architecture — shadow libraries to training data to commercial product — has been partially disrupted but fundamentally ratified. What remains is whether Congress, the courts, or market power will determine how much of the value created by creative work flows back to the people who made it.
