South Korea advances AI data and deepfake legislation

South Korea is advancing legislative changes to address legal gaps created by rapid artificial intelligence development. One proposed amendment to the Personal Information Protection Act would allow the use of original, non-pseudonymized personal data for AI training and performance improvement without explicit consent, provided the Personal Information Protection Commission approves on a case-by-case basis.

This amendment aims to resolve training-data bottlenecks but shifts decision-making authority toward the regulator. To qualify for use, the data must meet specific conditions: anonymization or pseudonymization must be insufficient, safeguards must be implemented, the purpose must serve a public interest or social benefit, and the risk of unfair infringement must be low.

Separately, lawmakers are moving to criminalize the creation, distribution, and possession of sexually explicit AI-generated material, even when it depicts fictional personas. This legislative push follows a court ruling in Goyang that acquitted a man accused of distributing deepfake images because the judge determined the victims were virtual rather than real people. Lawmakers argue that the difficulty in distinguishing virtual people from real individuals makes this a serious social issue.