Kakao signs agreement with Artificial Intelligence Safety Research Institute... Establishing a step-by-step verification system starting from the language model

Kakao [035720] is joining hands with the Artificial Intelligence Safety Research Institute to build an artificial intelligence (AI) safety evaluation system.

Kakao announced on the 29th that it signed a business agreement (MOU) with the Artificial Intelligence Safety Research Institute on the 28th to establish a joint AI safety evaluation and evaluation system.

Through this agreement, both companies plan to establish an objective and systematic safety evaluation basis for their AI models and AI agents.

Kakao provides models, agents, and infrastructure for evaluation, and the research institute is responsible for conducting evaluations and advancing methodology.

Cooperation proceeds in three stages.

First, we plan to start with a safety evaluation before and after distribution of the language model, expand the target to include multimodal language models and AI agents, and then jointly develop an evaluation tool for efficient verification.

Previously, Kakao conducted a joint safety evaluation with the research institute last year on its own language model 'Kanana-1.5-9.8B'.

Kim Se-woong, Kakao AI Synergy performance leader, said, "Through this cooperation, Kakao's AI model has an objective and strict verification system," and added, "We will prioritize safety and trust along with technological innovation."

Kim Myeong-joo, head of the Artificial Intelligence Safety Research Institute, said, "As AI technology becomes more sophisticated, a system that preemptively identifies and evaluates potential risks becomes more important. We will build a system that meets global standards."