en.Wedoany.com Reported - On June 24, Naver Cloud introduced the latest progress of its public sector AI platform "Clova Studio for Gov" at the Public AI Expo. Since the service launched in March this year, the platform has expanded to over 40 government departments and institutions, transforming from a mere AI development tool into a universal public AI infrastructure.
On the 24th, Naver Cloud Executive Director Jeong Ju-hwan stated at the briefing held in Hall 2 of KINTEX Ilsan that through the pan-government AI universal infrastructure project, the infrastructure is fully in place, and an environment has been established where public officials can directly use their own ideas and data to create AI agents and run them according to their workflows. He made these remarks while sharing the current status of the pan-government AI universal infrastructure.
Clova Studio for Gov is adding Software as a Service (SaaS) capabilities to its existing Platform as a Service (PaaS) role. Initially positioned as a tool to help public institutions develop AI transformation projects, it is now providing service-type functions that public officials can use without additional development. The translation service is a typical application: after uploading text, images, or documents, it can instantly translate them into over 27 languages, providing a Papago-like service within the administrative network without security concerns. A Markdown conversion function to transform public institution documents into AI-ready data is also planned for release in July.
The Agent Builder function is also being strengthened. It supports both workflow-based and agentic AI, and plans to add an MCP-linked client role. The RAG function will be upgraded to Graph RAG, multimodal RAG, and large-capacity file (over 100MB) processing. In terms of model support, the platform covers a range from HyperCLOVA X's HCX-GOV-Think 32B, 24B, VLM 24B, to LG AI Research Institute's K-Exaone 236B, Google Gemma 4 31B, Translation Gemma 27B, and GPT-OSS 120B, supporting diverse choices from sovereign AI models to global open-weight models.
Regarding the security system, the system is built within the administrative network, and query content and uploaded documents are not leaked or used for AI model training. The platform has passed the N2SF security verification hosted by the National Intelligence Service and KISA, and complies with the enforcement decree of the AI Basic Act. The cost structure is divided into two tracks: the Ministry of the Interior and Safety supports hackathons for enhancing public officials' AI capabilities and Agent Builder testing costs within the annual budget, creating a free trial environment; while departmental-specific AI projects use their own budgets to pay usage fees.
Functional upgrade plans for the second half of the year have been specified, including strengthening RAG functions such as Graph RAG, multimodal RAG, and large-capacity file (over 100MB) processing, while adding MCP-linked functionality to Agent Builder. The introduction of domestic NPUs and web search integration are also planned for support within the year, with web search limited to querying only public data.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
