Bae Hwan-guk, CEO of SOFTCAMP CO., LTD(258790), made these remarks on the 2nd at “SOFTCAMP Solution Day 2026,” held at the Fairmont Ambassador Seoul in Yeouido, Seoul. He emphasized that as the era of Agent AI—going beyond generative AI—gains momentum, we must move away from the traditional approach of unconditionally blocking AI use and adopt a new security paradigm of “Allow & Control.”
CEO Bae stated, “The era of telling people not to use AI is already over,” adding, “While the focus used to be on whether a network was closed or access was blocked, the focus is now shifting toward examining what kind of data is involved, who is using it, the scope of their permissions, and the level of AI access.”
SOFTCAMP presented the four core security principles of Trace, Control, Authenticate, and Verify as the foundation for this approach and introduced seven solutions based on these principles.These solutions track the creation and movement of data, control the leakage of confidential information to generative AI, authenticate the identities of humans and AI agents, and manage the entire lifecycle of the software (SW) supply chain.
CEO Bae personally shared his experience of creating a program using an AI agent to analyze internal sales schedules. He stated, “PATs (Personal Access Tokens) and secret keys must not be granted indiscriminately to AI agents,” adding, “The execution permissions of AI agents must be managed.”
SOFTCAMP showcased seven security solutions and consulting services centered on tracking, control, authentication, and verification—which it presented as the core principles of security in the AI era. Along with these, the company also unveiled two new solutions designed to support the secure use of AI agents.
The newly unveiled “Security365 Compliance Studio” is a visibility platform that, once integrated with Microsoft 365, assesses compliance with domestic security requirements—such as the Personal Information Protection Act. "Shield AI Gateway" is designed to manage not only external generative AI but also the invocation and permissions of internal AI agents, ensuring that security policies—rather than AI itself—determine the tasks AI is permitted to perform.
Next, Professor Jang Hang-bae of the Department of Industrial Security at Chung-Ang University took the stage and stated, “We are now in an era of defending against AI with AI,” adding, “We need a response system to enhance reliability—not only by ensuring the consistency of AI responses but also through de-identification and access control.”
He emphasized, “The key is not to block AI itself, but to classify sensitive data so that only necessary information is used safely.”
Professor Jang noted, “While security in the past involved blocking threats after they occurred, the paradigm in the AI era is shifting toward first identifying and classifying data that needs protection, and then controlling its use,” adding, “The key to the AI transformation era lies in the accurate identification and assessment of what needs to be protected.”
