Wednesday, April 1, 2026
Back to All Stories
AI & Publishing

Anthropic Settlement: Attorney Fee Cut Adds $112.5M to Author Pool — Claims Deadline Today

A significant development in the Bartz v. Anthropic class action settlement emerged ahead of today's March 30 claims deadline: plaintiffs' attorneys voluntarily reduced their fee request from $300 million to $187.5 million on March 23, 2026, adding $112.5 million directly to the author and publisher claimant pool. The total settlement fund remains $1.5 billion, covering approximately 500,000 books used without permission to train Anthropic's Claude AI models. Eligible authors are estimated to receive approximately $3,000 per qualifying work, with payments typically split 50/50 between authors and publishers. A final approval hearing is scheduled for April 23, 2026, with distribution calculations expected by June 11, 2026.

Legal settlement document on mahogany desk with calculator showing $1.5B and sticky note '$112.5M to authors' — Bartz v. Anthropic

Analysis

The Bartz v. Anthropic settlement has been the landmark AI copyright case since its announcement, but a development in the final days before the March 30 claims deadline deserves specific attention: the voluntary reduction of plaintiffs' attorney fees from $300 million to $187.5 million. In class action litigation, voluntary fee reductions of this magnitude are genuinely unusual. They typically occur when the settlement terms are generous enough that counsel can afford to take less, or when the optics of a $300 million legal fee on a $1.5 billion author compensation fund are sufficiently uncomfortable to motivate a recalibration.

The practical effect is meaningful. The $112.5 million that moves from the legal fee column to the claimant pool represents a 7.5% increase in the money available to authors and publishers. For individual claimants receiving approximately $3,000 per qualifying work, the difference is not transformative — but the signal it sends about the settlement's overall character is. This is a case in which the plaintiffs' counsel chose to prioritise the interests of the class over the maximum permissible fee, which is not the default outcome in high-value class action litigation.

The settlement's structural terms remain the most consequential aspect of the case for the industry's long-term trajectory. The $1.5 billion total represents the largest copyright recovery in US history, and the per-work payment of approximately $3,000 establishes a de facto market price for the use of a book in AI training — at least for the period covered by this settlement. That price will be cited in every future licensing negotiation between publishers and AI companies, and in every future lawsuit where the question of appropriate compensation arises.

What the settlement does not provide is an ongoing licensing framework. Authors whose works were used to train Claude before the settlement period are compensated; authors whose works may be used in future training rounds are not protected by this agreement. The Authors Guild and other advocacy organisations have been explicit about this gap: a one-time payment, however large, does not substitute for a sustainable licensing system that compensates creators on an ongoing basis as AI models continue to be trained and updated. The settlement is a milestone, not a resolution.

For the approximately 500,000 books covered by the settlement, the claims deadline of March 30 is today. Authors and publishers who have not yet filed should do so immediately. The final approval hearing on April 23 will determine whether the settlement proceeds as structured, and distribution calculations are expected by June 11.