In the near future, songs created by artificial intelligence in the style of BTS and BLACKPINK could be heard around the world. Global big tech companies that develop AI would earn revenue through usage fees, while Korean copyright holders who provided the underlying material would receive little meaningful compensation. As returns decline, investment in new planning and creative work would inevitably contract. The K-pop industry could then follow the path of Hong Kong cinema or retreat into a largely domestic market, as Japan’s idol industry has done. This is one plausible future that could be encouraged by the South Korean government’s policy of broadly granting immunity for the use of copyrighted works in AI training.
Recently, the National Artificial Intelligence Strategy Committee has been considering a proposal that would allow AI companies to use data for development without first obtaining consent from rights holders. The approach, often described as use first and compensate later, is now under active discussion. Even now, companies routinely cite trade secrets to avoid disclosing which copyrighted works they used, or how extensively, in training their AI models. When those whose property has been taken do not know what was used or in what amount, compensation is inevitably determined by the buyer. In effect, the proposal differs little from granting unconditional immunity for the use of copyrighted works in AI training.
If such a policy is implemented, works created through the blood, sweat, and toil of creators are likely to be treated not even as bulk commodities, but as clearance items sold at steep discounts. “My blood, sweat, tears, my last dance, take it all away. My body, my mind, my soul. I know they all belong to you. This is the spell that punishes me,” lyrics from BTS’ song “Blood Sweat & Tears,” offer an unsettling echo of that prospect.
It is understandable to agree that South Korea must not fall behind in the sweeping AI transformation. But if creators and copyright holders are expected to bear all the sacrifice, it is worth asking whether domestic AI companies can truly catch up with global big tech under those conditions.
In theory, it would be preferable for domestic data to be used only by Korean firms, but such a non-tariff barrier is unrealistic. Broad copyright immunity for AI training is far more likely to turn South Korea into an easy data mining ground for global big tech. AI developed by overseas frontrunners could train on Korean copyrighted works and improve output quality, including in the Korean language, more quickly than domestic firms can develop their own models. The result could be a double loss, weakening data sovereignty while still failing to close the AI gap.
Regulating AI training is not limited to copyrighted works alone. Global AI companies are also setting their sights on South Korea’s manufacturing data. Indiscriminate permission for AI training could ultimately amount to handing over core source data for industrial AI solutions directly to global big tech firms. The service sector is not immune to this risk either. When such concerns are raised, a common rebuttal is that stealing data, which are corporate assets, would obviously be prohibited. That response, in turn, raises a simple question. Why, then, is it considered acceptable to take copyrighted works.
The first step must be taken with care. The global standard is to seek a balance between AI development and the rights of copyright holders by putting safeguards in place. The European Union requires disclosure of AI training data, and Japan does not grant immunity when the interests of copyright holders are harmed. In the United States, courts rule on a case-by-case basis, but there is broad agreement that AI training that undermines the market for original works is unlikely to qualify as fair use.
To borrow an analogy from the late Joseon period, today’s policy, aimed at promoting modern industry, risks producing outcomes similar to the indiscriminate granting of mining rights to foreign powers. No matter how urgent the moment may seem, a thousand-mile journey cannot be made by looking only one step ahead.
Jong-Yeob JO jjj@donga.com
Most Viewed