South Korea’s current AI framework is being criticized for regulating risky systems after deployment while leaving no explicit ban on some uses considered unacceptable at the design, development, or training stage. The issue was raised at a National Assembly discussion on amendments to the Basic Act on Artificial Intelligence, according to a report by Maeil Business Newspaper.
From procedural safeguards to explicit prohibitions
The debate focused on systems that could turn prediction into punishment. Proposed areas for prohibition included AI that exploits vulnerable people or manipulates subconscious behavior, large-scale social scoring, facial-information databases built without consent, real-time remote biometric identification in public places, emotion recognition at work or school, and systems that infer sensitive information from biometric data.
A particularly serious concern involved predictive policing based only on profiling. Critics said such systems could classify people as potential criminals without a specific act or physical evidence, weakening the presumption of innocence and due process. Social scoring could create a similar structural risk by linking information collected in one area to disadvantages in employment, housing, finance, or public services.
The European Union was cited as a contrasting example. Article 5 of the EU AI Act prohibits certain practices, with related rules applying from February 2025 and potential penalties of up to 35 million euros or 7 percent of worldwide annual turnover. A National Assembly Research Service participant cautioned, however, that Korea should not simply copy overseas standards. The unresolved policy question is which decisions Korean society will refuse to delegate to AI, and how that boundary should be agreed through public debate. The discussion did not establish that a Korean ban has been adopted.
