“The current artificial intelligence (AI) race is focused on boosting performance, but going forward, using it safely will matter more than performance. In this context, if AI builds intelligence, Web3 (Web 3.0, an internet model where users directly own their data and distribute information) builds trust. By combining the two, we can create AI with trust and security built in.”

Raonsecure CEO Lee Sun-hyung, 56, said this in an interview with ChosunBiz at Raonsecure’s headquarters in Yeouido, Seoul, on the 10th of last month. Lee said, “In the era of agentic AI, it is important to control identity and behavior so that AI does not exceed the authority it is granted,” and added, “As with the national mobile ID, we should use Web3 technologies such as Blockchain-based decentralized identity (DID) to grant AI a verifiable identity, and leave records of delegated authority and behavioral history in a way that is hard to tamper with so they can be traced.”

Lee Sun-hyung, CEO of Raonsecure, sits for an interview with ChosunBiz at the company's Yeouido headquarters in Seoul on the 10th of last month. /Courtesy of RaonsecureLee Sun-hyung, CEO of Raonsecure, sits for an interview with ChosunBiz at the company’s Yeouido headquarters in Seoul on the 10th of last month. /Courtesy of Raonsecure

Founded in 2012, Raonsecure is an IT security and authentication company that provides security solutions and digital authentication services. After growing with a focus on biometric authentication (FIDO) and mobile and PC security, it absorbed its subsidiary Raon WhiteHat, which handled Blockchain-based digital authentication and security consulting, in Dec. 2023. Through this, it expanded its business into digital authentication, including DID, unified authentication, and USIM-based authentication. In particular, by participating in building national mobile IDs such as mobile government employee IDs, driver’s licenses, and veterans registration cards, it accumulated DID-based digital identity technology.

Recently, it has made AI a new growth pillar. Raonsecure held the “2026 SecuUp & Hackathon” on the 30th of last month and unveiled five agentic AI security solutions. ▲ “ONEAccess,” which manages the identities and privileges of AI agents and robots ▲ “ONEShield,” which blocks AI-targeted threats, including malicious prompts, in real time ▲ “ONETag,” which analyzes documents and automatically classifies security levels ▲ “ONESOAR,” which automates security operations and threat response through natural-language commands ▲ “ONEHacker,” an AI penetration testing solution. The company plans to link these products to build a security framework that spans all stages, from AI identity management to security operations and penetration testing.

Raonsecure posted sales of 63.751 billion won and operating profit of 2.678 billion won last year, up 2% and 36%, respectively, from a year earlier, continuing its growth streak. This year, it was newly included in the “KOSDAQ Rising Star,” selected by the Korea Exchange (KRX) based on an assessment of technology and growth potential.

After graduating from Hanyang University with a degree in chemical engineering, Lee entered the cybersecurity industry in 1995 by working in the in-house venture security organization at Mirae Industries. Since then, Lee has spent more than 30 years in the security industry, including as a founding member of first-generation domestic security company SoftForum. Lee assessed the changes AI will bring as even greater than when the internet emerged. Lee said, “AI has begun carrying out attacks that people couldn’t do,” but added, “Conversely, technologies that use AI to defend against and manage AI-driven attacks are also advancing rapidly.” The following is a Q&A with Lee.

—The government is emphasizing sovereign AI, including an independent AI foundation model project. Why is security sovereignty necessary?

“Security sovereignty does not mean excluding foreign technologies and using only domestic ones. We should use excellent overseas large language models (LLMs) and security technologies, while securing strategic autonomy to respond on our own when needed. In particular, we need the capability to respond to attacks tailored to Korea. There are attack patterns different from overseas, such as exploiting vulnerabilities in Hangul Word Processor (HWP) documents or bypassing cryptographic algorithms and authentication and security programs mainly used domestically. We must accumulate security logs and intrusion cases suited to these and establish response systems. At the same time, this security data must be thoroughly controlled. Ultimately, I see security sovereignty as securing Korea-specific security capabilities and safely managing and controlling the related data.”

Raonsecure unveils its AI and Web3-based digital trust strategy and AI security technologies at the 2026 SecureUp & Hackathon. /Courtesy of RaonsecureRaonsecure unveils its AI and Web3-based digital trust strategy and AI security technologies at the 2026 SecureUp & Hackathon. /Courtesy of Raonsecure

—As agentic AI spreads, what security threats grow?

“The biggest change is the speed and scale of cyberattacks. In the past, vulnerability discovery, writing exploit code, and building intrusion paths required substantial expertise and time, but now AI is rapidly automating this. As seen with ‘Claude Mythos,’ AI is going beyond vulnerability discovery to building attack paths and supporting actual attacks, shifting from ‘hackers using AI’ to ‘AI conducting attacks.’ Also, agentic AI can access internal systems with normal accounts and privileges, so if privileges are taken over, it can move like a legitimate user. In the end, beyond external perimeter defense, it becomes crucial to manage AI identity and privileges, and to track what actions were taken to the end.”

—Raonsecure has emphasized AI identity verification in AI security.

“There is a saying in the military: ‘If it is not identified, it cannot be controlled.’ I believe all control begins with identity verification. Agentic AI also needs to be given a unique identity, and we must be able to track what authority it was delegated and what actions it took. Using Blockchain-based DID, we can grant each AI a verifiable identity and tie it to privileges and execution history. ‘ONEAccess’ also starts from this concept. Just as people prove identity with a mobile ID, AI is given a digital identity and allowed to move only within permitted privileges. No matter how autonomously AI behaves, organizations and the state must not lose control over identity and privileges.”

—What differentiates Raonsecure’s AI security platform?

“Our AI security platform automates detection and response while holistically managing the identities, privileges, and behaviors of autonomously moving AIs. In addition, Raonsecure has, on top of its existing security capabilities, experience with Web3 technologies such as Blockchain and DID accumulated through the national mobile ID project. These two capabilities are connected to identity verification, privilege management, and tracking and control in the AI era. The five solutions unveiled this time do not stop at individual functions but interoperate within a single framework. By integrating AI autonomous defense and red teaming, identity verification, and behavior and data control, our strength is automating security while also managing AI’s own trust and safety.”

—You joined the Upstage consortium in the third and final phase of the independent foundation model project. What role will you play?

“I believe this participation is meaningful not only for Raonsecure but also for the domestic security industry. While LLM corporations had focused on performance and industrial application, there is growing recognition recently that ‘safety is as important as performance.’ Raonsecure will check the security vulnerabilities of Upstage’s LLM to enhance safety and take on the role of creating a Korea-specific security model by combining white-hat hackers’ attack and defense experience with security data. Ultimately, we aim to advance security and authentication technologies so the independent AI model can connect to and operate safely with real industrial data and systems, and to expand this into agentic AI across various industries, including finance and manufacturing.”

—What are your future plans?

“Going forward, we aim to build a ‘digital trust infrastructure’ that manages identities and privileges for agentic AI and physical AI, beyond people and devices. Our goal is to go beyond ‘Security with AI,’ which uses AI to strengthen security, and realize ‘AI with Security,’ with security built in from the start. Raonsecure’s direction is to enable safer use of all devices and services that incorporate AI, including drones, self-driving cars, robots, and home appliances.”

※ This article has been translated by AI. Share your feedback here.