The Telecommunications Technology Association (TTA) said on the 23rd that it granted the "Certification of AI Trustworthiness (CAT)" to KT's internally developed large language model (LLM) "Mit:um K 2.0 Base."
The Certification of AI Trustworthiness is a private voluntary certification program operated by TTA to address ethical and technical risks emerging with advances in artificial intelligence (AI) and to strengthen the competitiveness of Korea's AI industry. In 2025, the certification framework was advanced based on international standards (ISO/IEC 23894: AI risk management, 42001: AI management system, 38507: AI governance). TTA said this certification is the first case in which an international standard-based certification criterion has been applied to a large language model (LLM) with vast parameters and general-purpose capability.
KT's "Mit:um K 2.0 Base," which obtained the certification this time, has learned from large-scale Korean-language data and can perform a variety of Natural Language Processing (NLP) tasks such as classification, question-and-answer, summarization, generation, and transformation. It is a general-purpose AI model trained on Korea's language, culture, and sentiment. "Mit:um K 2.0 Base" is an open-source model with 11.5B parameters and is currently available on the Hugging Face platform.
Mit:um K 2.0 Base has learned from large-scale Korean-language data and can handle various Natural Language Processing (NLP) tasks such as classification, question-and-answer, summarization, generation, and transformation. It is an open-source model with 11.5B parameters and is currently available through the Hugging Face platform.
TTA said that through this certification it confirmed that, for the model in question, outputs and results by process—such as details on training data filtering, remediation of vulnerabilities through model red-teaming, and comparative tests of various trustworthiness evaluation benchmark performances—are being managed quantitatively and on an ongoing basis.
TTA President Son Seung-hyeon said, "Securing the trustworthiness of LLMs, which are the core of Generative AI, is an important foundation for corporations' responsible use of AI and for alleviating users' concerns," adding, "With this certification, corporations can manage AI risks systematically, and users can be more assured in using Generative AI services."