Ahead of the general elections, Kakao will embed watermarks in AI-generated images to prevent false and manipulated information.
According to industry sources on the 13th, Kakao has established policy and technical measures to prevent the creation and distribution of false and manipulated content.
Kakao introduced an invisible watermark for general users in Karlo, the image generation model developed by Kakao Brain. Even if users edit the images, the watermark remains intact.
The watermark feature was first implemented in the KakaoTalk channel's "Karlo AI Profile" function. It is also scheduled to be introduced within this month on "Karlo.ai," a web-based expert tool for image generation. Articles generated by AI for media outlets will display a phrase at the top indicating that the article was "written by an automatic generation algorithm" to ensure recognition.
Kakao is also simplifying the reporting process for false and manipulated information. When users report violations, they are connected to the Central Election Commission's reporting center. If deepfake videos or captured images from videos appear in public areas, measures such as deletion will be taken.
Additionally, Kakao is strengthening measures to prevent the spread of false information by posting banners on the Daum portal such as "Report False Facts and Defamatory Posts," "Report Misinformation and Rights Infringement," "Collection of Unfair Articles," and "Correction, Rebuttal, and Follow-up Reports."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
