Anthropic settles $1.5 billion AI book‑copyright lawsuit

U.S. authors led by Andrea Bartz, Charles Graeber and Kirk Wallace Johnson sued AI developer Anthropic, accusing it of downloading more than 7 million copyrighted books from pirate “ghost libraries” such as LibGen to train its Claude language model. A court ruled that while training the model falls under fair use, the unlawful copying and storage of the books violated copyright law. In July 2026 the parties reached a settlement worth $1.5 billion, with Anthropic agreeing to pay roughly $3,000 per book and to destroy all illegally obtained copies. The company had selected a “non‑destructive” scanning approach for processing the volumes. Political scientist Gleb Kuznetsov said the deal confirms that legally purchased books may be used for AI training, while illegal copies must be removed.

The settlement is viewed as a landmark for AI‑related copyright disputes and could influence how future AI firms source training data.