Judge approves Anthropic $1.5B copyright settlement; authors get ~$3,100 per work

On July 20, U.S. District Judge Araceli Martínez-Olguín granted final approval to Anthropic's landmark $1.5 billion class-action copyright settlement with authors whose works were used to train the Claude LLM. The settlement covers approximately 482,000 copyrighted works downloaded from pirate libraries (Library Genesis and Pirate Library Mirror). More than 91% of eligible authors and publishers have already claimed their shares, receiving approximately $3,100 per work after legal and administrative costs.

The underlying dispute centered on Anthropic's acquisition of millions of pirated books. Last year, Judge William Alsup ruled that using copyrighted works to train AI qualifies as fair use under U.S. copyright law, but found Anthropic violated copyright by maintaining a repository of 7+ million pirated books. This created potential liability in the hundreds of billions before settlement. Anthropic is required to destroy all copies of torrented and downloaded works and certify that pirated datasets were excluded from its commercial models.

This is the first major AI copyright settlement and sets a financial precedent: $3,100 per work is roughly four times the statutory minimum for ordinary copyright infringement ($750). However, the ruling does not establish binding precedent beyond this case—dozens of similar lawsuits continue against OpenAI, Google, Meta, and Microsoft. Architects should note: the fair-use finding (valid training) plus the settlement precedent (pay for pirated copies) may reshape how AI companies source training data, driving demand for licensed datasets and copyright-cleared corpora.