SEOUL, September 18 (AJP) - South Korea has chosen a middle path between Europe’s strict rules and U.S.’s hands-off stance in global race to govern artificial intelligence.
The Ministry of Science and ICT this week detailed the criteria for high-impact AI systems pertinent to the Act on the Development of Artificial Intelligence and Establishment of Trust" dubbed as the AI Basic Act due to take effect next January.
Seoul will be the second jurisdiction after the European Union to set statutory guidelines for AI, though with an emphasis on promotion over regulation during the early stage of the technology.
AI systems trained with cumulative computing power of 10^26 floating-point operations or more will be classified as high-impact — a threshold aligned with U.S. standards but looser than the EU’s 10^25 level. Companies deploying such systems will be required to adopt risk management plans, disclose training data and supervisors online, and provide users with clear advance notices through terms of service or interfaces.
Additional measures include watermarking AI-generated content, labeling deepfakes in ways easily recognizable to users, and marking outputs from generative AI for both human and machine readability.
While the act empowers regulators to impose penalties for noncompliance, the ministry indicated enforcement will be delayed. Fines are expected to be deferred for at least a year after the enactment.
“Promotion takes priority over regulation,” said Kim Kyung-man, AI policy director at the ministry. “We don’t intend to impose stricter rules ahead of other countries.”
The U.S. has taken a far looser approach, passing the “One Big Beautiful Bill Act” in July to block state-level AI regulation for a decade. Europe, by contrast, set the global precedent when its AI Act came into force in August 2024, imposing strict transparency, accountability, and anti-discrimination requirements on companies.
With its AI Basic Act, South Korea is seeking to balance light-touch regulation with global alignment as it pursues its ambition to become one of the world’s top three AI powers.
Copyright ⓒ Aju Press All rights reserved.
