MAIN ARTICLE
Twenty-one editions does not mean twenty-one authors
LTI Korea identified at least 21 English-language editions by Korean writers on the eligible works list. That unit matters: it is not a count of 21 authors, nor does it automatically include every original Korean edition. According to The Korea Herald, the institute has contacted the writers and intends to work with copyright organisations to support fair compensation, including legal action where necessary. The identification connects the dispute with particular works.
Examples include Han Kang’s Human Acts, translated by Deborah Smith; Kim Ae-ran’s My Brilliant Life, translated by Chi-Young Kim; and Hwang Sok-yong’s The Guest, translated by Chun Kyung-ja and Maya West. The report also names books by Kim Young-ha, Kim Soom, Shin Kyung-sook and Hwang Sun-mi. Crediting translators recognises the work involved in creating an English edition. It does not itself establish who holds which rights, a question depending on contracts and the relevant records rather than simply the names printed on a cover.
Acquiring books and training models are distinct questions
The June 23, 2025, order in Bartz v. Anthropic distinguished the training use before the court from downloading pirated copies for a central library. It accepted fair use for that training use but did not extend the same justification to the pirated library. The distinction should not become a blanket statement that all AI training, from any source and under every country’s law, is permitted. The particular conduct and scope of the ruling matter as much as the familiar phrase fair use.
Being on the settlement list therefore describes a work’s place in the dispute rather than supplying a technical record for every model. The list alone does not identify each training stage, every version of Claude or the origin of a particular response. That caution does not diminish authors’ interests. It helps keep questions about acquisition, storage and use clear enough to answer separately. A demand for transparency becomes more precise when it asks what happened to a work rather than assuming all those steps were identical.
A payment per work is not an automatic payment per person
The settlement totals $1.5 billion, with approximately $3,000 per work commonly cited. Authors Guild guidance explains that allocation can involve authors, publishers and other copyright owners.
The amount should therefore not be presented as a fixed payment to each individual creator, or as an identical additional award to everyone associated with a translation. A work and a recipient are different units. Translators’ entitlement and allocation still depend on the rights involved.
The official settlement website states that the claim deadline was March 30, 2026. It would therefore be misleading to describe newly identified writers as automatically entering a fresh open filing window. A listed work, a valid claim and a completed payment are different stages. Official notices and rights records govern the appropriate handling of a particular case. The supplied report does not confirm payment to each Korean author. The update concerns identification and support, not verified payment to every writer.
Bulk purchase requests raise a separate transparency issue
The Korea Herald says Kyobo Book Centre received requests from ten overseas companies, each seeking roughly 100,000 titles, and declined because it could not rule out AI training as the purpose. These were proposed purchases, not confirmation that a million books had been supplied to models. Some requesters denied a training purpose while naming data-related businesses as final recipients. The actual intended use remained a question for verification. Scale may justify inquiry without proving that every large order has an unlawful purpose.
Yes24 also reported one meeting with a US distributor about bulk purchases of secondhand books, followed by no further discussions. The Korean Publishers Association is surveying the industry before coordinating a response. Those developments concern transparency in transactions and the publishing sector’s ability to understand demand. They are separate from the American settlement. The available account does not establish that the purchase requests formed a supply route for Anthropic or connect every prospective buyer to the company. Similar concerns should not be turned into an unproven link between different events.
The case began before the Korean editions were identified
Three authors sued Anthropic in 2024 over copyrighted books downloaded from pirate sources. The 2025 fair-use order separated different uses, and the dispute over pirated copies proceeded to a $1.5 billion settlement. The official website confirms final court approval on July 20, 2026. LTI Korea’s identification of English editions adds a Korean perspective to that existing process, rather than creating a new case. Approval also differs from completed distribution of every payment. That chronology helps keep the legal milestones separate from the later work of identifying titles, establishing entitlement and handling individual claims.

















