Several authors awaiting payouts from Anthropic’s copyright settlement received unexpected emails this week: someone else was laying claim to their money.

Anthropic agreed to a $1.5 billion settlement in a class-action copyright lawsuit last year, after a judge ruled that using copyrighted material to train AI models constitutes fair use, but hoarding that material does not. The settlement received final approval in July 2026, and payouts have begun.

Under the terms, authors of nearly 500,000 titles will receive $3,000 for each unlawfully downloaded work. If a book remains in print with a traditional publisher, the payment is split 50-50 between the author and the publisher. If the book was self-published or rights have reverted to the author (out of print), the author receives the full amount.

The complication lies in determining who is entitled to what. On social media, numerous authors argue that publishers are claiming more than their fair share. Mystery writer April Henry shared that HarperCollins staked a claim on one of her books even though the rights had reverted at least 17 years ago, and on the same day, she received a notification that the system had classified HarperCollins as her “employer”—a relationship that never existed.

Victoria Strauss, who writes for the Writer Beware blog, grouped the complaints into two categories: publishers claiming payouts for works they no longer hold legal rights to, and publishers claiming 100% when they are only entitled to 50%. She was reluctant to attribute this to malice, noting that some publishers acknowledged the error and requested corrections from Anthropic. However, she also pointed out that the volume of reports surged in just two days with the same recurring errors, raising questions about whether this is a systemic issue rather than isolated mistakes. Mary Rasenberger, CEO of the Authors Guild, told The New York Times that she does not view this as publishers intentionally profiteering, but rather as the consequence of poor record-keeping and a convoluted settlement process.

Publishers are not the only ones staking claims. Strauss received complaints that some literary agents are also demanding a percentage, even though agents do not hold the rights to the works they represent. Author Courtney Milan directly called out this behavior.

To claim the full 100%, the rights reversion had to be finalized before August 10, 2022, the designated “download date” under the settlement. Milan and the Authors Guild have published guides to help authors dispute claims on their share.

The largest copyright settlement in AI history is encountering a familiar paradox. While the publishing industry has spent years criticizing AI for training on dirty data, the payout mechanism for the lawsuit itself hinges on author rights records that are anything but clean. Although there is no official finding of bad faith on either side, available evidence points to the process and record-keeping as the core vulnerabilities.