KT Cloud said on Thursday it will establish an AI safety cooperation framework with Kakao to implement safe AI services.

The companies will combine KT Cloud’s AI infrastructure platform capabilities with Kakao’s AI safety technology to build a foundation that lets customers develop and operate generative AI services with confidence.

As a first step, KT Cloud will link Kakao’s AI guardrail model, Kanana Safeguard, to its RAG-based generative AI service platform for public-sector customers, RAG Suite 2.0, introduced in April. Kanana Safeguard is an AI safety model specialised for the Korean language and Korea’s cultural and legal context. It supports safer use of generative AI services by detecting and blocking harmful content generation, inappropriate responses and prompt attacks.

The companies plan to provide a range of AI safety tools to KT Cloud customer environments in stages. These include not only guardrails but also a model stability evaluation system and an AI red-teaming system that diagnoses vulnerable areas based on various attack scenarios.

The companies will also expand cooperation into an operations platform field to integrate and manage AI safety tools, based on customer feedback and operational data collected in the KT Cloud environment.

KT Cloud CEO Bong-gyun Kim (김봉균) said that as generative AI spreads in earnest, building a foundation for responsible AI use, in addition to performance, is becoming an important task. He said the company will combine AI infrastructure and safety technology through cooperation with Kakao and support public and private customers so they can develop and operate AI services in a more reliable environment.