AI company Anthropic has agreed to a $1.5 billion settlement with authors and publishers after a court found it had illegally downloaded and stored millions of copyrighted books to train its AI models.
The terms of the agreement require a payment of $3,000 per work to an estimated 500,000 authors, making it the largest financial payout in the history of United States copyright litigation. The settlement is a pre-trial resolution and does not include an admission of wrongdoing by Anthropic.
This case is one of more than 40 similar lawsuits currently active between AI companies and copyright holders. Legal experts suggest the outcome could set a benchmark, potentially influencing other technology firms to seek settlements, court rulings, or direct licensing agreements for content used in AI training.
Final approval of the deal has been postponed by a judge, creating some uncertainty about its implementation. While the $3,000 per-work figure may become a standard for future cases, critics argue that the $1.5 billion total could create a high financial barrier that disadvantages smaller AI startups and favors large, well-funded corporations.
For Anthropic, the payment is considered manageable following a recent fundraising round that secured $13 billion, contributing to its $183 billion valuation.
The lawsuit also clarified a key legal distinction: while training an AI model on copyrighted material might be considered “fair use,” acquiring that material through illegal downloads from pirated sources is not protected by fair use principles.
