South Korea became the first in the world to regulate artificial intelligence. AI images are now mandatory to be labeled there
South Korea has become the first country in the world where laws regulating the operation of artificial intelligence systems have come into force. Images created with the help of AI must now be clearly labeled, and in important areas of activity, people will be obliged to control the work of computer systems, writes the BBC.
Photo: Jonathan Raa/NurPhoto via Getty Images
South Korean authorities state that the new laws are designed to strengthen public trust in artificial intelligence and ensure safety.
South Korea's Ministry of Science and ICT described this set of laws as the second of its kind in the world.
The European Parliament approved what it claims are the "world's first rules on AI" in June 2024. However, unlike the European Union, where a comprehensive set of laws on AI regulation will be introduced in stages until 2027, the South Korean equivalent came into force immediately in full.
Globally, disagreements still exist on how to regulate AI. The USA prefers a softer approach to avoid stifling innovation. China has also introduced a series of rules and proposed creating a body to coordinate global regulation.
What the laws entail
Under the new South Korean legislation, companies must ensure human control over so-called "high-performance" AI, including in areas such as nuclear safety, drinking water production, transportation, healthcare, and financial services (e.g., creditworthiness assessment and checks before issuing loans).
Other rules require companies to inform users in advance about products or services that use high-performance or generative AI, and to clearly label results obtained with AI. This includes deepfakes, which are difficult to distinguish from reality.
Authorities do not intend to immediately punish violators of the new legislation. Companies will be given a one-year grace period, after which administrative fines will begin to be imposed for violations of the law.
However, these fines can be significant: for example, the absence of generative AI labeling carries a fine of up to 30 million won (20,400 US dollars).
True, these amounts are small compared to the EU. In Europe, non-compliance can lead to fines ranging from 1% of a company's global turnover for minor violations to 7% for breaching prohibitions on the use of high-risk AI.
Meanwhile, many founders and executives of South Korean tech companies are frustrated, said Lim Chung-wook, co-head of Startup Alliance. "There is some dissatisfaction — why should we be the first to do this?" he noted.
According to Startup Alliance, some wording in the law is not precise enough, and companies may by default choose safer but less innovative approaches to avoid risks.
South Korean President Lee Jae-myung on Thursday called on politicians to heed the concerns of industry representatives and provide sufficient support to venture companies and startups.
"It is important to maximize the industry's potential through institutional support, while preventing anticipated side effects," he stated.
The Ministry of Science and ICT plans to create a platform with recommendations and a special support center for companies during the transitional period.
"In addition, we will continue to review measures to minimize the burden on the industry," a ministry official reported. He added that the authorities are considering extending the grace period if the situation in the industry requires it.
Comments