Judge rejects Anthropic’s $1.5 billion copyright settlement, demands clearer plan for authors
Anthropic’s legal troubles over training data aren’t over yet. Despite agreeing to what was touted as the largest copyright payout in U.S. history, a federal judge has rejected the proposed $1.5 billion class-action settlement, saying it was being pushed “down the throat of authors” and was “nowhere close to complete.”
The rejected deal would have paid roughly $3,000 each to around 500,000 writers and rights holders. While the court previously ruled in June that Anthropic’s use of copyrighted material to train its Claude large language model qualified as fair use, it also allowed separate claims to proceed over allegedly downloading books from websites such as Library Genesis.
In tossing the settlement for now, the judge flagged several major gaps:
– No complete list of works involved in the case
– No definitive list of affected authors and class members
– Insufficient process for notifying class members
– An inadequate claim form that failed to provide a clear, explicit choice
The court emphasized that class members often “get the shaft” once “money on the table” shifts focus away from robust protections for authors. To fix that, Judge Alsup instructed the attorneys to create a claim form that lets each class member clearly opt in or opt out of the settlement.
Key deadlines now in place:
– By September 15: Attorneys must submit a complete list of works covered by the case.
– By October 10: The court must approve the final list of works, a list of all class members, and the revised claim form for the settlement to move forward.
What this means for authors and AI companies
– Authors may gain stronger safeguards, clearer choices, and better notice before a settlement is finalized.
– Any future AI training data settlements will likely face stricter scrutiny on transparency, scope, and class member rights.
– Even with a fair use ruling for training, alleged piracy claims tied to how data was obtained remain a serious legal risk.
For now, Anthropic’s proposed payout is on hold until the court is satisfied that every affected author knows exactly what’s on the table—and has a genuine say in whether to accept it.
