As artificial intelligence (AI) transitions from a simple algorithmic competition to a phase where it controls factory equipment and influences battlefield decisions, a strategy for a 'national private AI ecosystem' that integrates key strategic areas of defense, manufacturing, and next-generation encryption and security technologies has been unveiled.
On September 7, the Seoul National University 10-10 Private AI Initiative and the Finance Ministry's Strategic Economic Advisory Group jointly hosted a forum titled 'Defense + Manufacturing + Security in the AI Era' at the Sangsan Mathematics Hall of Seoul National University, attended by key stakeholders from academia, government, military, and industry.
The forum focused on building a trust-based private infrastructure that utilizes data without exposure. The blueprint aims to elevate South Korea to a true 'AI superpower' and a leader in defense by integrating encryption computing technology as a foundational infrastructure, particularly in sectors where data leaks or contamination could have catastrophic consequences.
Park Young-sun, chair of the Finance Ministry's Strategic Economic Advisory Group and former Minister of SMEs and Startups, emphasized in her opening remarks, "The past few years have been the 'first act' of AI, where we confirmed what generative AI can do. Now, the 'second act' asks where we connect AI and how we will make the real world move with it. AI must transform manufacturing, which in turn enhances defense competitiveness, while security must maintain the trust of this entire AI ecosystem." She added that national competitiveness in the AI era stems not from a single technology but from a strategic economy that weaves together semiconductors, power, data, and security.
Kim Joo-han, Vice President for Research at Seoul National University, highlighted in his welcoming address that the nation's competitiveness in the AI era relies on the 'safe utilization of sensitive data without leaks.' He noted that global leaders in defense AI are emerging, with companies like Palantir becoming market focal points. "AI competitiveness is not determined solely by faster and larger computing power. It requires the systematic acquisition of high-quality data in the field and ensuring safety through thorough verification without external leaks, which will ultimately shape the nation's industry and future competitiveness," he stated.
Kim also mentioned that Seoul National University has been focusing on core technologies that can safely handle sensitive national and industrial data by breaking down the barriers between mathematics, AI, policy, and industry. He pledged that the university would continue to lead in establishing standards for 'trustworthy AI' and in deriving policy tasks and creating new innovation ecosystems.
During the defense session, experts diagnosed the challenges faced by K-defense in hardware-centric exports and suggested a shift towards a software and AI-driven ecosystem.
Son Dae-kwon, head of the Logistics AX Center at Inha University and former Army Logistics Command Chief, warned in his presentation on 'Defense AX Strategy and the Future of the Defense Industry' that while K-defense is looking at a backlog of orders worth 100 trillion won, the export of single hardware products will soon face limitations due to global demand and geopolitical constraints. He cited an example from the Israel-Iran conflict, where an AI system identified 15,000 targets in just two weeks, stating, "The rules of the battlefield have already been reshaped by software and data."
Son emphasized the need to transition from being a 'hardware supplier' to an 'ecosystem designer' that provides independent operating systems and solutions. He argued for the establishment of a 'Defense AX Full Stack' ecosystem that integrates infrastructure, data, models, and services to break the bottleneck between weapon systems and defense AI. He called for a 'Export 2.0' structure that continues to provide AI updates and intelligent maintenance after weapon exports, locking in allied nations within our ecosystem. He also advocated for a four-stage growth ladder (R&D → PoC government procurement → new security fund → global scale-up) to support this transition.
Won Jun-hee, president of the Defense SMEs and Startups Association, presented on creating an ecosystem for defense SMEs and software innovation for AI exports. He pointed out that while 88% of domestic defense companies are SMEs, they account for only 18% of total sales, with 96% of exports dominated by a few large corporations, highlighting the subcontracting structure that favors large companies.
Won explained four policy solutions and expected outcomes for independent entry into the global value chain (GVC) without being dependent on large system integrators. He stated, "Defense SMEs and software that obtain international standards and certifications can be supplied universally to weapon systems worldwide, regardless of the success of specific large contracts," and urged the establishment of a dedicated R&D track for SMEs (70% allocation), the creation of an international standards and certification hub, a GVC business certification system by growth stage, and the assignment of intellectual property rights of government service software to developers.
He concluded that if these institutional innovations are completed, the sales share of defense SMEs could expand to over 30%, and South Korea could localize and export 56 trillion won (5%) in the global 1,114 trillion won materials and parts market, ultimately raising the national defense export scale to 60 trillion won, achieving the goal of 'prosperous nation, strong military.'
The manufacturing session addressed strategies for innovation in productivity within the nation's foundational industry through AI implementation and overcoming limitations.
Yoon Jong-pil, head of the Manufacturing AI Research Center at the Korea Institute of Industrial Technology, pointed out that the current AI adoption rate in South Korea's manufacturing sector is only 4.0%. He identified key bottlenecks, including a low AI exposure rate of 0.26 compared to finance and services, the fact that 68.9% of small manufacturing companies have an initial investment capacity of less than 100 million won, lack of compatibility with legacy systems, and the challenge of hiring specialized personnel, which affects 49.2% of companies.
Yoon introduced advanced demonstration cases, such as vibration signal-based tilt diagnosis for semiconductor CMP process cleaning equipment, defect inspection for medical catheter tubes, and multi-robot monitoring, stating, "The key to successful manufacturing AI lies in redefining field problems as problems that AI can solve."
He added that manufacturing AI does not operate solely on a single advanced model; rather, a collaborative system involving 'field data experts' with domain knowledge, 'AI model experts' who design neural networks and characteristics, and 'AI system experts' who integrate with existing factory systems (MES, ERP, etc.) must work together to create economic value through process optimization and predictive maintenance.
As AI becomes more deeply integrated into defense and manufacturing, security emerges as a critical prerequisite. The security session highlighted the need to establish homomorphic encryption (HE) and primary data protection technologies (PETs) as foundational elements of national infrastructure.
Cheon Jeong-hee, a professor in the Department of Mathematical Sciences at Seoul National University and head of the Crypto Lab, pointed out in his presentation on 'Encrypted Computing and AI Security' that a critical flaw in existing security systems lies in the 'in-use data' that is decrypted when data is computed in memory. He explained that the tacit knowledge of manufacturing processes and classified defense data have been actively withheld from sharing due to the potential for external leaks, which could threaten the existence of companies and the nation.
Cheon presented the homomorphic encryption search technology (HEracles) that performs SQL searches and AI learning without decryption, advocating for the establishment of protective and usable encrypted computing as the 'default' for national AI infrastructure. He noted that homomorphic encryption is no longer just a theoretical technology; it has already been demonstrated in defense experimental projects, police data, and the Korean Manufacturing-X platform. He stressed that while AI's power comes from connecting data, national trust arises from not exposing that data. He called for the design of a security layer that encompasses defense, manufacturing, and public safety as a common national foundation to be rapidly disseminated as procurement standards.
Ryu Geun-kwan, an emeritus professor in the Department of Economics at Seoul National University and former head of the Statistics Korea, presented on 'National Data Safety Linkage System (Hub & Spokes) and Security,' proposing a method for integrating statistical and administrative data. He stated, "To establish evidence-based precise policies, longitudinal and cross-sectional data integration is essential. Blocking data to avoid risks is a misguided solution that eliminates the evidence for policy." He suggested a 'hub and spoke' model where the central statistical registry (hub) connects with various departments and agencies (spokes) without exposing original data.
Ryu particularly urged the immediate resumption of R&D projects based on internationally recognized privacy-enhancing technologies (PETs: homomorphic encryption, differential privacy, synthetic data) for safe linkage. He noted that the pilot project for homomorphic encryption by Statistics Korea has already been included in the 2023 UN PET guidelines and recognized as a model innovation case in reports from Germany's Federal Ministry of Labor and Social Affairs and the UK's Department for Science, Innovation and Technology.
He warned that repeating the information strategy plan (ISP) while having already completed global demonstrations of the encrypted linkage system is a serious waste of time and a setback for national competitiveness. He called for the immediate transition of the R&D project, which faced budget cuts, to a main project centered on national data processing and urged for expedited pilot construction to safeguard both data sovereignty and policy quality in the AI era.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.