Exhibition of Multimodal and Safeguard Models
Experience Model Features and Functions in Demonstration Area
On August 5, Kakao announced its participation in the 'Asia-Pacific Economic Cooperation (APEC) 2025 Korea Global Digital·AI Forum and Corporate Exhibition,' where it will introduce its artificial intelligence (AI) technologies.
This forum, held as a side event of the APEC Digital·AI Ministers' Meeting, takes place at the Sheraton Grand Hotel in Songdo, Incheon. More than 10 domestic and international IT companies, including Kakao, are participating in the corporate exhibition held outside the main forum venue.
At its exhibition booth, Kakao will showcase its proprietary lightweight multimodal language model 'Kanana-1.5-v-3b' and the AI guardrail model 'Kanana Safeguard.' Visitors can experience the features and functions of these models through demonstration areas within the booth.
Released as open source on July 24, Kanana-1.5-v-3b is a lightweight multimodal language model capable of processing not only text but also image information. It is described as having strong instruction-following performance for understanding user intent and the ability to comprehend images in both Korean and English.
For example, if a user attaches a photo of a location and asks for an explanation of where it was taken, the model can respond with, "This photo features the Gwanghwamun area in Seoul," providing answers based on its recognition of local information and Korean culture.
Kanana Safeguard, a guardrail model for verifying the safety and reliability of AI services, is built using Kakao's proprietary language model as its core technology. It features performance optimized for the Korean language by utilizing a custom-built dataset that reflects Korean language and culture. The model consists of three types, enabling detection of harmfulness and risk according to the type of risk. In May, Kakao became the first Korean company to release a guardrail model as open source.
A Kakao representative stated, "We aimed to provide an opportunity for visitors to directly experience the current status of Kakao's AI technology development through the exhibition space," and added, "We will continue to work on invigorating the domestic AI ecosystem and strengthening competitiveness based on ongoing AI technology development."
Meanwhile, as a separate event during the forum, a session hosted by global big tech company Meta featured Kim Kyunghoon, Kakao's AI Safety Leader, as a panel speaker. Kim led a discussion on the value of open-source AI.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
