"To Entrust Real Work to AI"…Raonsecure Focuses on 'Safety' Over Performance

Lee Sunhyung, CEO of Raon Secure, presented strategies for digital trust necessary in the AI and Web3 era during the keynote speech at the "2026 Secu-Up & Hackathon." Provided by Raon Secure

"Can we really entrust 'real work' involving sensitive information to AI?"

This question resonated throughout "SecureUp (SQUP) 2026" hosted by Raonsecure at Fairmont Ambassador Seoul in Yeouido, Seoul, on September 30. As the race to improve artificial intelligence (AI) performance accelerates, Raonsecure pointed out that it is not only about how smart AI is, but also how safely it can be controlled in order to handle actual work.

The theme of SecureUp, marking its 14th year in 2026, was "make Agentic AI fun and secure." At this event, Raonsecure introduced an array of technologies and products required for agentic AI to perform work on behalf of humans, including identity verification, authority management, data protection, breach response, security operations, and prompt attack defense.

Raonsecure is expanding its accumulated expertise in digital authentication to the AI domain. The company's decentralized identity (DID) technology has been applied to the national mobile ID used by around 45 million people in South Korea. This event focused on expanding traditional identity verification and authorization control systems—originally designed for people—to AI agents and robots.

Lee Sunhyung, CEO of Raonsecure, emphasized in his keynote speech that it must be possible to verify who the AI acted on behalf of, which system it accessed, and what actions it took. He explained that for AI to independently take over human work, not only execution capabilities, but also robust control and accountability frameworks must be established first.

As a solution, he proposed Web3-based AI identity technology. By combining digital identity, delegated authority, audit and traceability, and data sovereignty, the idea is to foster trust in AI activities. This is essentially an expansion of DID principles, already applied to mobile IDs, into the AI space.

He stated, "If AI continues to enhance its performance and execution capabilities, then Web3 provides the trust framework that makes those actions reliable." He added, "It must be possible to confirm whose authorization was used and what actions were taken, in order to ensure accountability and ultimately control."

He further emphasized, "Ultimately, the key is not to restrain the autonomy of AI, but to make it possible to utilize AI more freely and safely within a trustworthy scope."

The five agentic AI security solutions unveiled at the event embody this strategy in concrete products. "OneAccess" assigns unique identities to AI agents and robots, and manages their permissions, delegation relationships, and usage history. Even if AI accesses systems on behalf of humans, it enables verification of who granted which authority and what actions were performed.

Raonsecure also introduced "OneHacker," an AI-based penetration tester that verifies vulnerabilities and attack vectors in AI systems from an attacker's perspective. Other solutions included "OneTag," which automatically classifies documents by security level through contextual analysis, and "OneSOAR," which automates security operations such as account/privilege management and threat response using natural language commands.

"OneShield," an AI guardrail that prevents threats arising from malicious prompts and jailbreak attacks in AI input and output processes, is also part of Raonsecure's agentic AI security framework.

Raonsecure not only aims to provide these products as individual solutions, but also to integrate them into a single security framework. From the stage of managing AI identity and authority to data classification, vulnerability verification, automation of security operations, and prompt attack defense, Raonsecure envisions protecting each step in the entire process as AI performs real work.

Taejin Kim, Chief Technology Officer (CTO) of Raonsecure, commented, "We didn't just release individual AI-related products, but designed the entire AI operation environment as a single pipeline," adding, "The key differentiator is that each technology is organically linked, protecting everything from AI security to operations."

The event also featured "Raoni," Raonsecure's first humanoid AI employee. Raoni demonstrated various identity and authentication scenarios, such as issuing its own employee badge after an authentication process and cross-verifying customers' mobile IDs. This showcased a structure in which AIs and robots, like humans, are given identities and permissions and are controlled to act only within authorized boundaries.

The event also included sessions on innovations in practical work ecosystems using AI coaching, verification services based on mobile IDs, and future strategies for agentic AI. Meanwhile, in the finals of the "Blockchain & AI Hackathon" held the same day, 10 teams that passed the preliminaries competed by showcasing services utilizing mobile IDs.

This content was produced with the assistance of AI translation services.