Illicitly Collected K-pop Photos Fuel Widespread Distribution of AI Models | Be Korea-savvy

Illicitly Collected K-pop Photos Fuel Widespread Distribution of AI Models


This image, captured from CivitAI.com, shows generative AI models.

This image, captured from CivitAI.com, shows generative AI models.

SEOUL, June 28 (Korea Bizwire)Artificial intelligence (AI) models, which utilize machine-learned photos of K-pop girl group members collected illicitly, are being extensively distributed online.

CivitAI, a platform for sharing generative AI models, hosts numerous AI datasets derived from learned photos of popular K-pop girl groups such as Le Sserafim.

The published files consist of AI models capable of generating deep fake images using Stable Diffusion, an open-source AI program for image blending provided by Stability AI last year.

Stable Diffusion generates images based on user-entered prompts, utilizing “models” as an extensive collection of machine-learned images.

Low-Rank Adaptation (LoRA), a model used for generating deep fake images of real individuals, occupies a size no larger than tens to hundreds of megabytes. It is relatively easy to create and widely shared on platforms like CivitAI.

The issue lies in the potential exploitation of these AI models for profit and the production of pornography.

YouTube hosts numerous so-called AI lookbooks, showcasing AI-generated female models dressed in various outfits and costumes.

Certain users even sell pornographic content created by AI on Patreon, a sponsorship platform.

Legal experts emphasize that generating profit through content created with AI models that violate the portrait rights of celebrities or influencers may result in criminal or civil punishment.

Kevin Lee (kevinlee@koreabizwire.com)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>