As the Supreme Court hears an appeal regarding copyright infringement damages related to the National Institute of the Korean Language's corpus project, attention is focused on how the court's decision may impact the trading of AI training data.
On October 30, the Korean Publishers Association (KPA) stated that the settlement reached with Woongjin Booksen in 2023, which served as the basis for calculating damages in the appeal, reflects the public interest of the project and the unique circumstances of dispute resolution.
The KPA clarified that this settlement amount should not be used as a general standard for determining the value of AI training data.
The KPA emphasized that the agreement was made after considering various factors, including the project's aim of building public language resources, the early resolution of the dispute, and future usage conditions. They argued that applying this specific settlement amount as a general pricing standard for current AI training data transactions is inappropriate.
As the market for AI training publishing data begins to take shape, the KPA noted the need to distinguish between past settlement agreements aimed at resolving disputes and current licensing transactions. There is growing concern that past settlement amounts could be cited as the de facto 'market price' due to the lack of publicly available reference cases.
In particular, since the payment criteria from the previous settlement were considered a major basis for calculating damages in the appeal, the KPA insists that the circumstances and scope of the agreement must be adequately reflected. They warned that if these amounts are accepted as 'standard compensation' in the data acquisition processes of public projects or private AI companies, it could distort the new market from the outset.
The KPA also highlighted that the 2023 settlement allowed for the possibility of additional adjustments and that the licensing fees should vary based on the purpose and duration of use, the scope of data provided, the number of companies utilizing it, and the potential for third-party use. They stressed that these differences must be fully considered before applying the previous settlement amount as a benchmark for other transactions.
To foster a healthy new market, the KPA proposed principles of prior licensing and fair compensation. They further stated that the public interest objectives of public projects and the responsibilities of companies conducting these projects or acquiring data must be clearly distinguished.
The KPA concluded, "Through licensing transactions on behalf of publishers, authors and publishers will receive fair compensation, while AI companies will obtain legally compliant and reliable high-quality data, establishing a trading order that allows both the publishing and AI industries to develop together."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.