Kakao is joining hands with Google DeepMind to strengthen transparency for AI-generated content.
Kakao said on the 20th that it signed a technology partnership with Google DeepMind to identify AI-generated content. Under the collaboration, Kakao will apply Google DeepMind's digital watermarking technology "SynthID" to its in-house AI model "Kanana."
SynthID is a technology that inserts a digital watermark, which is difficult for users to notice, into images, text, audio, and video so they can verify whether the content was generated by AI. Because it allows the source of generation to be traced without degrading content quality, it is drawing attention as a technology that can curb the spread of deepfakes and misinformation.
Starting in the second half of this year, Kakao plans to apply SynthID to the image-generation model "Kanana Collage" and the video-generation model "Kanana Kinema." It will first introduce it to "Kanana Template," which creates short videos in KakaoTalk using images. Kakao plans to apply an invisible watermark along with the existing logo display so that whether content was generated by AI can be identified even if the output is edited or reprocessed.
Kakao will be the first among Asian companies to apply SynthID to AI-generated content. The company plans to accelerate efforts to build a responsible AI ecosystem beyond legal obligations, in step with the implementation of AI-generated content labeling.