# Authors Push Back as Publishers and Agents Claim Larger Shares of Anthropic Settlement
Anthropic's legal settlement with authors over AI training data has triggered a brewing dispute over how the money gets distributed. Authors are pushing back against what they characterize as overreaching claims by publishers and literary agents seeking disproportionate cuts of the payout.
The tension centers on a fundamental question: who owns the rights to literary works used in training large language models, and therefore who deserves settlement compensation when those rights are violated. Publishers argue they control distribution and licensing rights, entitling them to significant portions of any settlement. Literary agents, who represent authors, also claim substantial shares based on their contractual relationships with writers.
Authors themselves contend they are the actual rightsholders and creators whose work was used without consent or compensation. They argue that publishers and agents are inserting themselves as middlemen in disputes that fundamentally concern the original creators.
The Anthropic settlement emerged from legal challenges brought by authors including Sarah Silverman, Michael Chabon, and John Grisham, who alleged the AI company trained its models on copyrighted books without permission or payment. While terms of the settlement remain partially under seal, the payout structure has become contentious as various parties claim entitlement to portions of the settlement pool.
This conflict reflects broader tensions in the publishing ecosystem around AI training. Publishers have taken mixed stances on generative AI, with some seeking licensing deals with AI companies and others pursuing litigation. Authors have largely been skeptical of both AI companies and their traditional publishers, viewing both as potential threats to their livelihoods.
The distribution dispute also highlights a gap in how settlement frameworks allocate funds in digital-era copyright cases. Traditional book publishing settlements typically flowed through publishers first, with authors receiving their contracted percentage afterward. But settlements specifically for unauthorized AI training raise fresh questions about whether that model applies when the core claim involves an author's intellectual property rather than a publisher's commercial interest.
Several author advocacy groups have begun mobilizing to ensure direct payments reach writers. They argue that without clear mechanisms to bypass intermediaries, authors could see settlement proceeds significantly diluted by publisher and agent commissions.
The case carries implications beyond this single settlement. As more copyright litigation involving AI training data proceeds through courts, the precedent set for distribution could shape how future settlements compensate creators. Tech companies training on unlicensed content, publishers defending their interests, and authors fighting to maintain control of their work will all watch how this particular distribution battle resolves.
The outcome could determine whether authors receive meaningful compensation for unauthorized use of their work, or whether traditional publishing intermediaries effectively capture most settlement value despite being secondary parties to the core dispute.