A federal judge on Monday approved Anthropic’s $1.5 billion out-of-court settlement with authors and publishers who objected to the company downloading books from pirate libraries to develop its Claude model. Anthropic collected more than seven million book files to build a permanent research library.
Terms of the settlement
- The settlement covers 482,460 eligible works.
- Eligible works were valued at about $3,000 each before fees.
- Funding for the payments will continue through 2027.
The agreement represents the largest copyright recovery in U.S. history.
Earlier ruling: training deemed fair use
A prior court ruling treated Anthropic’s model training as fair use. Given that finding, the $1.5 billion payment addresses only the downloads themselves, not the act of training the model or the model’s subsequent outputs.
Financial context and significance
According to figures cited in coverage, Anthropic was valued at $965 billion in May and reported a revenue run rate that had crossed $47 billion. On those bases, the $1.5 billion settlement equals roughly 0.16% of the stated valuation and about twelve days of revenue at the cited run rate. Market observers might view that as immaterial to the company’s finances, while copyright lawyers characterize the recovery as historically significant.
Why this matters
The settlement shows that copyright disputes do not necessarily block the development of frontier AI: they can be resolved through large payouts and multi-year funding arrangements. At the same time, the size of the award and the court’s distinction between downloading and training create legal and ethical precedents likely to influence future litigation and data‑acquisition practices for firms building large language models.
Next steps
Payments under the settlement will be funded through 2027, providing an organized schedule for compensating eligible rights holders. The decision and settlement details are expected to shape legal strategies and corporate procurement policies in similar cases going forward.
