Anthropic Nears Final Approval of Landmark AI Copyright Settlement
Key Points
- Anthropic agreed to a $1.5 billion settlement over alleged copyright infringement.
- Authors will receive $3,000 for each qualifying work used to train Claude.
- Nearly 100,000 claims have been filed; less than 0.5% of class members opted out.
- The company pledged to destroy all pirated copies and certify future models are free of pirated content.
- Judge William Alsup initially withheld approval, prompting an extensive outreach effort.
- A searchable settlement website provides details and a database of qualifying works.
- The court will consider the final approval motion in late April.
Anthropic is close to securing final court approval for a historic settlement that resolves claims that its Claude AI model was trained on pirated books. Nearly 100,000 authors have filed claims, and the company has agreed to pay a total of $1.5 billion, with $3,000 allocated to each qualifying work. The settlement includes a certification that no pirated content will be used in future Claude releases and a commitment to destroy existing pirated copies. The court is set to consider the final approval motion in late April, marking a significant milestone in AI‑related copyright litigation.
Background of the Lawsuit
The case, Bartz v. Anthropic, alleges that Anthropic illegally downloaded more than 500,000 copyrighted works from the pirated sites Library Genesis and Pirate Library Mirror to train its Claude artificial‑intelligence chatbot. Plaintiffs argue that this use constitutes copyright infringement, and the lawsuit has become the first major AI‑related case to reach a settlement.
Scope of the Settlement
Anthropic has agreed to a settlement valued at $1.5 billion. Under the agreement, authors whose works were used will receive a payout of $3,000 per qualifying work. Nearly 100,000 claims have been filed, and less than 0.5% of class members have opted out of the settlement.
The company also pledged to certify that no pirated content will be used in any released version of Claude and to destroy all pirated copies of the class members’ works.
Judicial Oversight and Outreach
U.S. District Judge William Alsup initially withheld approval, expressing concern that qualifying authors might not have sufficient time to join the settlement. In response, Anthropic launched an extensive outreach program that included traditional mail, email campaigns, social‑media outreach, and engagement with professional and informal author and publisher networks. A settlement website now hosts a searchable database of qualifying works and provides full details, including key documents and dates.
Upcoming Court Action
The court is scheduled to hear Anthropic’s motion for final approval in late April, with the last day for class members to file a claim set for the end of March. The settlement has garnered support from several writers’ groups and organizations, positioning it as the largest copyright payout in history.
Implications for the AI Industry
This settlement marks a pivotal moment for the AI sector, highlighting the legal risks associated with training models on unlicensed material. By agreeing to compensate authors and to eliminate the use of pirated content, Anthropic sets a precedent that may influence how other AI developers source training data.