I Wasn’t Sure I Wanted Anthropic to Pay Me for My Books—I Do Now (www.wired.com)

Anthropic’s recent $1.5 billion settlement to compensate authors and publishers whose books were used without permission to train its large language model, Claude, marks a pivotal moment in the AI copyright debate. This legal resolution, prompted by a judge’s summary judgment ruling the training data usage as piracy, introduces real monetary stakes into a conversation that has until now been largely theoretical. Authors could receive at least $3,000 per book, spotlighting the broader question of whether AI companies building trillion-dollar enterprises owe royalties for the valuable, coherent, and comprehensive knowledge embedded in books—an indispensable training resource for LLMs. The settlement underscores a critical tension in AI development: while copyright law’s “fair use” doctrine currently allows companies to exploit copyrighted works if their models’ use is deemed “transformational,” this legal framework struggles to fit the novel realities of AI training. Industry leaders and policymakers, including figures like former President Trump and AI czar David Sacks, have argued paying authors could jeopardize US competitiveness against China in AI, framing fair use as a necessary exception for national security. However, critics note that systems akin to the complex royalty tracking in the music industry could be adapted for AI training data, suggesting the “too hard to implement” rationale is more a strategic stance than a technical barrier. Ultimately, the Anthropic case propels the AI community to confront the fairness and sustainability of relying on creative works without adequate compensation. It not only challenges the ethical boundaries of how AI models are built but also raises cultural concerns over the decline of deep reading and authorial recognition in an era where AI-generated summaries risk supplanting the rich, nuanced engagement that books provide.