Report from Korea JoongAng Daily
In Brief – South Korea’s Ministry of Culture, Sports and Tourism has released an English version of its guide on the application of the fair use exception in South Korea’s Copyright Act to the training of generative AI models. The 75-page guide outlines four legal factors used to assess fair use, including purpose, type of work, amount used, and market impact. AI training may qualify as fair use in cases where it creates new value and does not negatively affect existing markets. The new guide also provides examples of uses that may not qualify as fair use under current copyright rules, such as models “creating no new value nor purpose,” those trained “excessively on a particular work,” systems that try to substitute the original market or bypass paywalls, systems that summarize full-text news articles without permission, or image generators trained on scraped stock photos. The ministry said it plans to discuss the guide internationally, including at the World Intellectual Property Organization.
Context – The debate over AI training and copyright law is global, but a series of lawsuits in US courts will likely have the greatest impact on the resolution. There were a pair of dueling decisions from US District judges in last June on the key “fair use” question, one defended training as “fair use” while the other one heavily criticized that claim, proposing the novel theory of “indirect substitution” through which AI systems nullify the fair use defense by creating massive volumes of cheap content that, while not copies of the originals, are “similar enough to compete with the originals and thereby indirectly substitute for them,” harming copyright holders. In the EU, with its AI Act, the copyright section of the General-Purpose AI Code of Practice has proven contentious, as have proposals in the UK and Australia to make their copyright regimes more conducive to AI training, while Japan’s training-friendly regime is garnering greater pushback from copyright holders. Last December, an Indian government panel recommended a very different approach that rejects broad fair use for training, calling for mandatory royalty payments for AI training on Indian creators’ content through a centralized licensing pool.