Warnings have emerged that entrusting artificial intelligence (AI) control in the defense sector to foreign companies could halt wartime operations entirely. In an era where AI is deployed across command, reconnaissance, target analysis, and other operational functions, access denial is not merely a service disruption\u2014it could lead to operational failure.<\/p>\n
At the third seminar of the “AX Strategy Forum” hosted by the Korea University Future Growth Institute on the 27th at the Korea Press Center in Jung-gu, Seoul, Bae Seong-jun, Executive Director of External Strategy and Partnerships at Naver Cloud, made this point in a presentation titled “A Sovereign AI Strategy for the Defense of the Republic of Korea.” He stressed that what matters most in defense AI is not just AI performance, but who controls the AI.<\/p>\n
The case of Anthropic blocking overseas customers’ access to its AI models at the direction of the U.S. government is a prime example. If AI access is cut off during wartime, operations themselves could come to a standstill. This is the so-called “kill switch” risk. The implication is that when AI is directly tied to national security, its availability must not be determined by the decisions of external companies or foreign governments.<\/p>\n
The presenter assessed that AI is already transforming the speed of decision-making in modern warfare. He cited the U.S. operation against Iran, “Operation Epic Fury,” where Palantir’s Maven Smart System and xAI’s Grok were used to support strikes on more than 2,000 targets within 96 hours.<\/p>\n
France is also integrating AI and cloud infrastructure that it can control into its defense sector, centered on Mistral AI and Dassault Syst\u00e8mes. The moves by the United States and France demonstrate that securing AI sovereignty has emerged as a core agenda item in global defense strategy.<\/p>\n
Limitations of General-Purpose AI and the Need for Defense-Specialized Models<\/p>\n
Applying general-purpose AI directly to defense has clear limitations. The South Korean military possesses unique knowledge that general AI models cannot easily access\u2014not only military acronyms and specialized terminology, but also unit organization, terrain, operational procedures, and rules of engagement (ROE). General-purpose AI models struggle to understand this context and face constraints operating on closed networks disconnected from the external internet.<\/p>\n
Accordingly, a “defense-specialized model” trained on additional military data is necessary. However, a specialized model alone is insufficient. The knowledge an AI has learned is static information. On an actual battlefield, situations change in real time. Enemy artillery may have moved 30 minutes ago, new drone footage may arrive, or logistics supply routes may be altered.<\/p>\n
Defense AI must go beyond merely answering based on previously learned knowledge\u2014it must be able to check real-time data and make judgments. This is where AI agents come in. An AI agent is not simply an AI that answers questions; it is an AI that searches for necessary information across multiple systems, analyzes the situation, and then connects to the next task.<\/p>\n
Harness, Data Fabric, and Dedicated Region<\/p>\n
One of the key concepts Naver Cloud presented is the “Harness.” A harness is a control mechanism that sets permissions and rules so that AI does not act arbitrarily, and connects it to the necessary military systems. Rather than allowing AI to freely access all defense data, it defines who can view what information and what tasks can be performed.<\/p>\n
It connects to C4I (command and control systems), surveillance and reconnaissance imagery, and unit databases, while restricting access scope according to security classification and user permissions. It also records what information the AI accessed and what judgments it made. Translating commanders’ experience and decision-making criteria into rules that AI can understand is also critical.<\/p>\n
For defense AI to function properly, data must also be connected. Currently, data is dispersed across the Army, Navy, Air Force, and Joint Chiefs of Staff. Rather than moving all data to a single location, Naver Cloud proposed a “Data Fabric” that keeps data in place at each unit while connecting it for use when needed. This approach binds data from different locations into a single connected network, enabling AI to find the information it needs.<\/p>\n
An “ontology” is applied on top of this so that AI can understand the relationships and meanings between data points. An “OMNI model” that understands not only text but also photos, video, and audio is also utilized. The vision is for AI to comprehensively grasp battlefield conditions by analyzing drone footage, terrain and weather information, and unit status together.<\/p>\n
The core of defense sovereign AI is operating independently within the defense network\u2014from infrastructure to AI services. A “Dedicated Region” is built on the military’s closed network, which is disconnected from the external internet. A dedicated region involves separately building AI servers and cloud infrastructure for a specific military branch or agency, enabling AI operations even in environments isolated from the outside.<\/p>\n
Naver Cloud’s “defense sovereign AI full stack” refers to building and controlling the entire defense AI system as one integrated framework\u2014from infrastructure to AI models, data and platforms, and AI agents.<\/p>\n
Field Engineers and Co-Evolution<\/p>\n
Naver Cloud also argues that the way AI is actually applied to the military must change. A representative approach is the FDE (Forward Deployed Engineer). Instead of AI developers building systems in offices and then delivering them, FDEs go directly into military field sites and build the necessary AI together with military personnel.<\/p>\n
Field engineers first understand the actual work processes and data flows, then develop one or two core functions. Subsequently, the system is continuously refined by incorporating requirements that emerge as the military uses it directly. Executive Director Bae introduced examples from Palantir and others in the U.S., emphasizing that defense AI must not end as a one-time system build. What is needed is “co-evolution”\u2014the long-term integration of military expertise and AI technology developing together.<\/p>\n
Following the presentation, a panel discussion on defense AI sovereignty and data utilization was held, moderated by Lee Seong-yeop, chairman of the AX Strategy Forum and professor at Korea University. Lim Hee-joo, professor of business administration at Seowon University, said, “Securing independent AI technology will become the core foundation for protecting South Korea’s defense and security.” Jeon Hyun-seok, professor of aerospace systems engineering at Sejong University, emphasized that AI application is also needed in the defense space sector for managing space objects such as satellites.<\/p>\n
The discussion included Nam Han-gyeol, attorney at Kim & Chang; Shim Seung-bae, Ph.D. at the Korea Institute for Defense Analyses; Ryu Hyuk-seon, professor at KAIST College of Business; Hong Ki-hyun, director of the Sejong DX AI Research Institute; Min Dae-ki, professor of business administration at Ewha Womans University; Kwon Kyung-yong, research fellow at Korea University; Jang Jae-man, director of the Air Force AI New Technology Convergence Center; and Kim Man-soo, former Air Force brigadier general. Chairman Lee said, “The defense AX strategy is expanding into substantive discussions spanning industry, academia, and the security field,” adding, “I hope this forum contributes to establishing defense AI security and enhancing the competitiveness of the defense industry.”<\/p>\n","protected":false},"excerpt":{"rendered":"Warnings have emerged that entrusting artificial intelligence (AI) control in the defense sector to foreign companies could halt…\n","protected":false},"author":2,"featured_media":135218,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[623],"tags":[9651,68664,68666,68668,68669,42678,28429,68667,68665,42376,657,944,22372,4570],"class_list":["post-135217","post","type-post","status-publish","format-standard","has-post-thumbnail","category-naver","tag-anthropic","tag-bae-seong-jun","tag-dassault-systemes","tag-data-fabric","tag-dedicated-region","tag-defense-sovereign-ai","tag-fde","tag-harness","tag-korea-university-future-growth-institute","tag-mistral-ai","tag-naver","tag-naver-cloud","tag-palantir","tag-xai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts\/135217","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/comments?post=135217"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts\/135217\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/media\/135218"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/media?parent=135217"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/categories?post=135217"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/tags?post=135217"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}