Korea to Extend AI Labeling Duty to YouTubers, Other Users

The proposed amendment seeks to address a practical challenge within AI regulation. The current law requires AI companies to visibly label deepfake content, but the technology struggles to determine what qualifies as a deepfake according to the legal definition. This technical difficulty means companies often label all AI-generated content visibly to ensure compliance, creating an overly broad application of the rule.

By shifting the labeling responsibility to users who produce or distribute deepfake content, the government aims to circumvent this technical hurdle. Officials believe users are better positioned to judge whether their AI-generated content is indistinguishable from reality and thus requires a watermark. This approach acknowledges that AI's current interpretive capabilities regarding human perception are limited.

Government officials indicated that extending the watermarking duty beyond January is becoming necessary because deepfake misuse is spreading into criminal activities such as voice phishing and sex offenses, as well as the broader content industry. This suggests a recognition that the existing regulatory framework, focused solely on AI developers, is not effectively addressing the proliferation of unlabelled deepfakes.