A federal judge on Monday gave final approval to Anthropic's $1.5 billion class action settlement with authors and book publishers, concluding the largest copyright class action ever certified and delivering what the plaintiffs' law firm called the "largest known copyright recovery in history"1,4,5.
Judge Araceli Martinez-Olguin signed off on the deal, writing that the settlement will provide "meaningful relief". Judge William Alsup, who issued preliminary approval last year, has since retired2.
The settlement covers an estimated 500,000 works and will pay approximately $3,000 per work, shared among the authors and publishers who hold rights to them. Bloomsbury, publisher of the Harry Potter series and home to authors Sarah J Maas and Susanna Clarke, said it had 14,087 titles listed within the settlement3.
The case traces back to a lawsuit first filed by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused Anthropic of training its AI models on copyrighted books. The court found that Anthropic had built its training library from two sources: books it purchased and scanned, and books it downloaded from pirate sites including Library Genesis and Pirate Library Mirror. Judge Alsup ruled that training an AI model on copyrighted text counts as fair use — a decision widely seen as a turning point for the AI industry — but found that downloading books from pirate sites was illegal on its own terms.
ANALYSIS The fair use ruling on AI training is the more consequential legal outcome embedded in this settlement.
The settlement faced opposition from some authors. In May, a group of authors fought to block the deal, arguing that lawyers' fees were too high and per-work payouts were too low. A handful of authors attempted to opt out past the deadline, hoping to file separate lawsuits seeking higher damages. Only 350 authors opted out in total.
Despite the record payout, many authors and creators do not view the outcome as a win, given the court's fair use ruling on AI training.
The settlement resolves one case but does not address the broader legal question of using copyrighted works to train AI models. ANALYSIS The bifurcated outcome — fair use for training, liability for piracy — draws a sharper line between licensed or purchased training data and material obtained through unauthorized channels.
Anthropic recently announced a strategic partnership with AMD to deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs ctx. The settlement's financial burden arrives as the company scales its infrastructure commitments.