# Authors Push Back Against Publisher Claims in Anthropic Settlement Dispute
Authors are challenging what they describe as overreaching claims by publishers and literary agents in disputes over settlement payments from Anthropic, the AI safety company behind Claude. The conflict reveals tensions between content creators, intermediaries, and AI companies over how compensation for training data should be distributed.
The dispute centers on lawsuits filed against Anthropic alleging that the company used copyrighted works to train its large language models without permission or compensation. Anthropic reached settlement agreements, but disagreement erupted over who gets paid and how much. Publishers argue they hold rights to the copyrighted material and deserve the lion's share. Authors counter that they own the underlying work and publishers are attempting to capture compensation that rightfully belongs to creators.
This mirrors similar conflicts playing out across the AI industry. OpenAI, Google, and other model developers face dozens of copyright litigation cases from authors, publishers, and news organizations. The Authors Guild, represented in some of these disputes, has pushed for settlements that benefit individual creators rather than corporate intermediaries. The guild views publisher claims as attempts to double-dip, collecting compensation for works while authors receive nothing.
The mechanics of the dispute are straightforward but contentious. When a publisher holds rights to distribute a book, questions emerge about whether those rights extend to AI training data use. Publishers argue that their licensing agreements grant them authority to negotiate and collect payments. Authors argue that publishers license distribution rights, not training or derivative work rights, making these separate claims. Some publishers have attempted to claim settlement funds on behalf of all their authors without direct authorization.
Anthropic's settlement structure matters here. The company structured payments to address multiple claimant categories, but the language leaves room for interpretation about priority and distribution methodology. This ambiguity has created space for competing claims. Unpacking these disputes requires examining the original licensing terms between authors and publishers, which vary widely across contracts spanning decades.
The stakes extend beyond individual payouts. How this resolves sets precedent for future AI training settlements. If publishers successfully extract the majority of settlement funds, it creates a model where content creators receive minimal compensation while intermediaries profit. If authors prevail, it establishes that original creators maintain claims over their work's use in AI training, regardless of publishing arrangements.
The publishing industry has experienced consolidation, with major houses like Penguin Random House, Hachette, Simon and Schuster, and HarperCollins controlling significant backlists. This concentration gives them negotiating power in settlement discussions. Individual authors typically lack equivalent leverage.
The dispute also raises questions about settlement transparency. Authors pushing back say they lack visibility into what publishers claim or how funds were allocated. Some authors report learning about settlements through news reports rather than direct notice from their publishers or agents.
This conflict will likely influence how future AI companies negotiate settlements. Platforms may insist on transparent distribution mechanisms or require direct author consent before publishers collect on their behalf. Alternatively, some companies might structure settlements to pay authors directly, bypassing publisher intermediaries entirely.
The broader pattern shows that copyright disputes in AI training extend beyond lawsuits against companies. They now encompass internal conflicts about how compensation distributes among copyright holders, licensees, and intermediaries. These disputes will shape whether AI training settlements benefit creators or primarily pad corporate bottom lines.