Anthropic $1.5B Settlement Stalls Over Fee Details
Key insights
- Judge declined final approval at the May 14 hearing, requesting supplemental detail on attorney fees and lead-plaintiff compensation.
- Claims rate reached 92.77% covering 447,576 works, pushing per-work payout down from $4,876 to roughly $3,000.
- Settlement remains pending and is not rejected; supplemental filings from both sides are still forthcoming.
Why this matters
The per-work payout drop illustrates a structural tension in large AI copyright settlements: high participation rates, while legally favorable, dilute individual recoveries and could fuel objections that undermine final approval. For AI founders and legal teams, the court's scrutiny of attorney fees in a case of this scale sets a precedent for how future AI training data settlements will be structured and reviewed. If approved, the $3,000-per-work figure becomes the de facto market price for retrospective training data licensing, giving every foundation model lab a concrete liability benchmark for unlicensed book ingestion.
Summary
Anthropic's $1.5 billion book-copyright settlement cleared nearly every hurdle except the last one: a federal judge declined final approval at the May 14 fairness hearing, citing insufficient detail on attorney fees and lead-plaintiff compensation before she'll sign off.
Judge Araceli Martínez-Olguín's pause isn't a rejection, but it keeps the settlement in limbo while both sides prepare supplemental filings. The participation rate has surged to 92.77%, covering 447,576 works, which sounds like a win until you see the math: higher claimant volume pushed the per-work payout down to roughly $3,000 from the original $4,876 projection.
Essentially: (Anthropic, plaintiff authors) are waiting on a judge who wants the fee structure scrutinized before the deal closes.
- Claims rate of 92.77% is unusually high for a copyright class action, suggesting strong author mobilization around AI training grievances.
- Per-work payout dropped ~38% from initial projections as participation outpaced fund assumptions.
- Court's focus on attorney fees signals scrutiny of whether counsel's cut is proportionate to the author recovery.
If approved, this will stand as the largest AI training copyright settlement on record, setting a cost-per-work benchmark every foundation model developer will need to price into future data decisions.
Potential risks and opportunities
Risks
- Prolonged court review could embolden objecting authors to push for a higher per-work floor, potentially unraveling the settlement and forcing Anthropic back into active litigation.
- If the settlement collapses, Anthropic faces renewed discovery exposure on the specific composition of its training datasets, which could surface liabilities beyond the book corpus covered here.
- Other foundation model developers (Google, Meta, OpenAI) watching this proceeding may face copycat filings using the $3,000 per-work figure as a floor demand before any settlement talks begin.
Opportunities
- Legal technology platforms specializing in class action claims administration (Epiq, Kroll) gain credibility as essential infrastructure for AI copyright settlements at this participation scale.
- Authors' rights organizations and literary agents can use the 92.77% claims rate as evidence of strong collective action potential, strengthening leverage in ongoing negotiations with other AI developers.
- AI developers with proactive licensing programs (e.g., those already negotiating with publishers) can use the $3,000 per-work benchmark to frame licensing deals as cheaper than litigation, accelerating deal flow.
What we don't know yet
- What specific attorney fee percentage is being requested, and how does it compare to the $1.5B gross fund versus the net author recovery?
- Whether the supplemental filings have a court-set deadline, and when a revised approval hearing might be scheduled.
- How lead-plaintiff compensation amounts compare to the average class member payout of ~$3,000, which the judge may view as disproportionate.
Originally reported by arstechnica.com
Read the original article →Original headline: Anthropic's $1.5B Copyright Settlement Hits Snag as Federal Judge Declines Final Approval, Requests More Detail on Attorney Fees