Kakao is joining hands with Google DeepMind to introduce invisible watermarks into its self-developed artificial intelligence models, aiming to improve content transparency and prevent the misuse of deepfakes.

Kakao said on Tuesday it signed a technology partnership with Google DeepMind to ensure transparency for AI-generated content. Under the agreement, Kakao will apply the digital watermarking solution SynthID to its in-house AI model, Kanana. It is an early adoption case globally and the first among Asian companies.

SynthID is a technology developed by Google DeepMind that verifies whether content is AI-generated by embedding watermarks in images, text, audio and video that cannot be detected by human senses. It can trace the origin of generated content without degrading quality, and is cited as a key alternative to prevent the misuse of false information.

Kakao plans to apply SynthID in phases from the second half of this year to its image model Kanana Collage and video model Kanana Kinema.

It will first be introduced to the KakaoTalk Kanana Template feature. Kanana Template is a feature that uses images exchanged on KakaoTalk to create short videos using AI, based on the Kanana Kinema model. Previously, only a logo was displayed on generated videos, but once SynthID is applied, the output will be able to be perfectly identified as AI-generated even if it is edited or reprocessed.

The move is also a pre-emptive response timed for the full introduction of a system requiring the labeling of AI-generated content, alongside the enforcement of the AI Basic Act this year. Kakao plans to boost trust in its platform further by introducing invisible watermarking technology beyond legal obligations.

Kim Kyung-hoon (김경훈), Kakao's AI safety leader, said, "As AI technology rapidly spreads into daily life, the importance of ensuring transparency and reliability for AI-generated content is also growing." He added, "Starting with the application of SynthID, we plan to continue technical cooperation with Google in the area of responsible AI and provide safe AI services that users can use with confidence."

Pushmeet Kohli (푸시미트 콜리), vice president for science and strategic initiatives at Google DeepMind, said, "Effective collaboration is essential to identify AI-generated content." He added, "By expanding the application of SynthID technology to more platforms through partnerships with industry-leading companies, we are taking an important step toward building a more transparent digital ecosystem for everyone."