[E-Daily Reporter Ho-Jun Kim ] Innotena Co., Ltd. (CEOs Kim In-soo and Yoon Young-ho), a company specializing in AX, announced on the 30th that it has signed a Memorandum of Understanding (MOU) with AI deep tech company Motif Technologies Co., Ltd. (CEO Lim Jeong-hwan) to collaborate on an AI-based intelligent document centralization business for use in closed-network environments.

The two companies plan to combine Innotena’s integrated information security brand “Mpower” with Motif Technologies’ proprietary AI foundation model and on-premises inference technology to jointly build a platform that enables the use of generative AI without connecting to external networks. In the public, defense, financial, and manufacturing sectors, it is often difficult to use external cloud-based AI services due to security regulations. Even when organizations attempt to build their own systems, GPU costs and access control issues have been cited as major obstacles.
This is precisely where the strengths of the two companies converge. Innotena’s “Mpower EZis-C” is a document centralization solution that consolidates work documents scattered across individual PCs into a central repository and blocks local storage, thereby already bringing an organization’s document assets and access control system together in one place. By integrating an AI model—designed entirely in-house by Motif Technologies, from the architecture to the tokenizer and inference kernel—with this solution, organizations can perform knowledge search and analysis on internal documents without the need for external integration.
Key areas under consideration for collaboration include: △ designing a reference architecture that integrates the Mpower product suite with AI models; △ developing a closed-network knowledge search (RAG) engine that accurately reflects user-specific access permissions; △ developing a context-based sensitive information classification model; and △ verifying on-premises inference performance and standardizing hardware requirements. The RAG engine is designed to filter permissions during the search phase and maintain citation sources and audit logs, ensuring that content from documents the user is not authorized to view does not appear in AI responses. The sensitive information classification model expands the scope of classification to include indirectly identifiable information and trade secrets—which are difficult to detect using traditional pattern-matching methods.
The two companies are exploring ways to apply Motif Technologies’ models tailored to customer scale and use cases. Motif Technologies’ “Motif 3” model employs a Mixed Expert (MoE) architecture—activating only 13.2 billion parameters per token out of a total of 314 billion—to enhance inference efficiency. They plan to progressively define future collaborative projects.
Kim In-soo, CEO of Innotena, stated, “The document assets and authorization system we have built through document centralization will serve as the most secure knowledge base in the AI era.” He added, “By combining these with Motif Technologies’ proprietary AI model, we will establish an environment where generative AI can be used with confidence—even without connecting to external networks—as the standard for the public and financial sectors.”
Lim Jeong-hwan, CEO of Motif Technologies, said, “We will collaborate with Innotena to ensure that AI can be used with confidence even in environments where security is critical.”