Authors Challenge Publishers and Agents Over Share of Anthropic Deal

Authors say they were excluded from negotiations as intermediaries press to claim a portion of the AI company's payout

Publishers and literary agents are moving to secure slices of a settlement announced with Anthropic, the makers of advanced language models, and a growing number of writers are pushing back. Authors say intermediaries are negotiating on their behalf without permission and are demanding direct compensation or the right to opt out of any licensing arrangement that uses their work.

The dispute centers on a recent agreement intended to resolve claims that large language models were trained on copyrighted books without authorization. Under the deal, an industry-wide fund and new licensing terms would compensate rightsholders for past and future uses. Publishers and agencies argue they represent authors and should administer parts of the payout; many writers counter that publishing contracts do not clearly give those intermediaries authority to approve AI uses or to accept money in lieu of author consent. The disagreement has turned what was billed as a settlement into a contested negotiation.

Writers' groups and individual authors have begun mobilizing. Several have sent letters and launched online campaigns demanding transparency about any monies distributed and an itemized accounting of how decisions are made. One novelist, who asked to speak on background, said she felt blindsided and accused her publisher of negotiating "behind closed doors." Others are weighing legal action to reclaim control over licensing decisions for their texts.

Legal scholars and industry observers say the dispute raises unresolved questions about contract law and representation. A law professor who has tracked AI litigation noted that whether publishers can bind authors depends on the precise language of each contract, and that courts have issued mixed signals in related disputes. "In many agreements, the rights granted to publishers were not drafted with machine learning training in mind," the professor said, adding that agents may face scrutiny over whether they fulfilled duties to their clients when consenting to large-scale uses.

The outcome could reshape how future AI deals are structured. If publishers and agents prevail in claiming centralized control, authors could see payments routed through intermediaries and limited options to refuse. If authors succeed, companies developing models may have to negotiate directly with vast numbers of individual writers or fund broader collective mechanisms that distribute proceeds more widely. Some industry attorneys expect the parties to seek mediation before filing new suits, but they warn that protracted litigation remains possible.

For now, talks continue behind closed doors while advocacy and legal posturing intensify. Authors say they will press for a seat at the table or for mechanisms that guarantee a direct share of any settlement funds; publishers and agents insist they are acting to secure the best outcome for the industry. Both sides acknowledge that the resolution will influence how creative work is treated in machine learning for years to come.