Raon Secure 'One Shield' Acquires Confirmation of AI Products and Services in Korea
Real-time Analysis of Generative AI Input/Output and Blocking Prompted Injection and Jail Attack
사진 확대 Raon Secure has obtained a confirmation of artificial intelligence products and services with its artificial intelligence (AI) security solution "One Shield." Its strategy is to target the public and financial AI security market with technology that blocks personal information leakage through Generative AI and malicious prompt attacks.
On the 14th, Raon Secure announced that AI-safe guardrail solution One Shield has obtained "Artificial Intelligence Product and Service Confirmation" from the Korea Artificial Intelligence and Software Industry Association (KOSA).
KOSA's artificial intelligence product and service confirmation system is a system that checks whether AI technology has been actually applied to products or services in accordance with the Basic Artificial Intelligence Act. A confirmation letter is issued by technically examining the learning, reasoning characteristics, and functional contributions of AI applied to the product.
One Shield is a security solution that analyzes data between users, Generative AI, and AI agents in real time. In accordance with the security policy set in advance by companies or institutions, information entered into AI and results generated by AI are inspected and allowed or blocked.
It also has a function of preventing information leakage that may occur during the process of using a Generative AI. When executives and employees enter personal information into AI, they automatically identify it and anonymize it or mask it. It also filters out harmful content based on Korean-specific models.
It also responds to attacks that cause AI to bypass existing security policies or restrictions by entering malicious commands. It is a method of detecting and blocking "prompt injection" that inserts malicious commands into prompts or "Jailbreak" attacks that neutralize safety devices set in AI.
With the acquisition of this confirmation, Raon Secure plans to expand its one-shield business to public and financial institutions. As more and more AI agents perform actual tasks beyond information search and answer generation, the demand for security to manage not only AI's input/output but also access rights and actions is expected to increase.
Lee Jung-ah, CEO of Raon Secure, said, "As AI agents shift beyond the use of Generative AI to perform actual tasks, new security threats such as leakage of sensitive information, malicious prompts, and policy bypass are also growing."