Anthropic to pay $1.5B over pirated books in Claude AI training — payouts of roughly $3,000 per infringed work

Anthropic, the company behind Claude AI, has agreed to pay $1.5 billion to settle a class-action lawsuit over the use of pirated books in training its AI models. The settlement, led by authors Andrea Barta, Charles Graeber, and Kirk Wallace Johnson, involves payouts of around $3,000 per infringed work. Anthropic will delete the infringing data but may not be required to retrain its models. This case sets a new precedent for data liability in AI development and highlights the risks of using pirated content. If models trained on pirated data face legal challenges, developers may need to start over with licensed datasets, potentially increasing demand for GPUs.